E-VO.guide

@database "e.guide"
@node MAIN
@title "Amiga E-VO v3.9.3"

                +-----------------------------------------------+
                |                                               |
                |               Amiga E-VO v3.9.3               |
                |          Compiler for The E Language          |
                |     Originally By Wouter van Oortmerssen      |
                |           Updates By GRIO 1999-2003           |
                |          and Darren Coles 2022 - 2026         |
                |                                               |
                +-----------------------------------------------+

Contents:

         0. compiler and introduction
          @{"  A. introduction  " link CH_0A }
          @{"  B. the distribution  " link CH_0B }
          @{"  C. installing/using the compiler  " link CH_0C }
          @{"  D. changes / history / new features  " link CH_0D }
          @{"  E. additional information  " link CH_0E }

         1. format
          @{"  A. tabs,lf etc.  " link CH_1A }
          @{"  B. comments  " link CH_1B }
          @{"  C. identifiers and types  " link CH_1C }

         2. immediate values
          @{"  A. decimal (1)  " link CH_2A }
          @{"  B. hexadecimal ($1)  " link CH_2B }
          @{"  C. binary (%1)  " link CH_2C }
          @{"  D. float (1.0)  " link CH_2D }
          @{"  E. character  " link CH_2E }
          @{"  F. strings ('bla')  " link CH_2F }
          @{"  G. lists ([1,2,3]) and typed lists  " link CH_2G }
          @{"  H. lisp-cells (<a|b>)  " link CH_2H }

         3. expressions
          @{"  A. format  " link CH_3A }
          @{"  B. precedence and grouping  " link CH_3B }
          @{"  C. types of expressions  " link CH_3C }
          @{"  D. function calls  " link CH_3D }

         4. operators
          @{"  A. math (+ - * /)  " link CH_4A }
          @{"  B. comparison (= <> > < >= <=)  " link CH_4B }
          @{"  C. logical and bitwise (AND OR NOT << and >>)  " link CH_4C }
          @{"  D. unary (SIZEOF PSIZEOF ARRAYSIZE OFFSETOF ` ^ {} ++ -- -)  " link CH_4D }
          @{"  E. triple (IF THEN ELSE)  " link CH_4E }
          @{"  F. structure (.)  " link CH_4F }
          @{"  G. array ([])  " link CH_4G }
          @{"  H. float operator (!)  " link CH_4H }
          @{"  I. assignments expressions (:=)  " link CH_4I }
          @{"  J. sequencing (BUT)  " link CH_4J }
          @{"  K. dynamic memory allocation (NEW)  " link CH_4K }
          @{"  L. unification (<=>)  " link CH_4L }
          @{"  M. pointer typing (::)  " link CH_4M }
          @{"  N. swap (:=:)  " link CH_4N }
          @{"  O. quick compare (==)  " link CH_4O }
          @{"  P. increment decrement etc (+= -= etc)  " link CH_4P }

         5. statements
          @{"  A. format (;)  " link CH_5A }
          @{"  B. statement labels and gotos (JUMP)  " link CH_5B }
          @{"  C. assignment (:=)  " link CH_5C }
          @{"  D. assembly mnemonics  " link CH_5D }
          @{"  E. conditional statement (IF)  " link CH_5E }
          @{"  F. for-statement (FOR)  " link CH_5F }
          @{"  G. while-statement (WHILE)  " link CH_5G }
          @{"  H. repeat-statement (REPEAT)  " link CH_5H }
          @{"  I. loop-statement (LOOP)  " link CH_5I }
          @{"  J. select-case-statement (SELECT)  " link CH_5J }
          @{"  K. increase statement (INC/DEC)  " link CH_5K }
          @{"  L. void expressions (VOID)  " link CH_5L }
          @{"  M. memory deallocation (END)  " link CH_5M }
          @{"  N. generate an error or warning (FATAL/WARN)  " link CH_5N }
          @{"  O. code and data sections (SECTION)  " link CH_5O }
          @{"  P. Negative conditions/loops (IFN/ELSEIFN etc)  " link CH_5P }

         6. function definitions and declarations
          @{"  A. proc definition and arguments (PROC)  " link CH_6A }
          @{"  B. local and global definitions: scope (DEF)  " link CH_6B }
          @{"  C. endproc/return  " link CH_6C }
          @{"  D. the 'main' function  " link CH_6D }
          @{"  E. built-in system variables  " link CH_6E }
          @{"  F. default arguments to functions  " link CH_6F }
          @{"  G. multiple return values  " link CH_6G }
          @{"  H. function values  " link CH_6H }

         7. declaration of constants
          @{"  A. const (CONST)  " link CH_7A }
          @{"  B. enumerations (ENUM)  " link CH_7B }
          @{"  C. sets (SET)  " link CH_7C }
          @{"  D. built-in constants  " link CH_7D }

         8. types
          @{"  A. about the 'type' system  " link CH_8A }
          @{"  B. the basic type (LONG/PTR)  " link CH_8B }
          @{"  C. the simple type (CHAR/INT/LONG)  " link CH_8C }
          @{"  D. the array type (ARRAY)  " link CH_8D }
          @{"  E. the complex type (STRING/LIST)  " link CH_8E }
          @{"  F. the compound type (OBJECT)  " link CH_8F }
          @{"  G. initialisation  " link CH_8G }
          @{"  H. the essentials of the E typesystem  " link CH_8H }

         9. built-in functions
          @{"  A. io functions  " link CH_9A }
          @{"  B. strings and string functions  " link CH_9B }
          @{"  C. lists and list functions  " link CH_9C }
          @{"  D. intuition support functions  " link CH_9D }
          @{"  E. graphics support functions  " link CH_9E }
          @{"  F. system support functions  " link CH_9F }
          @{"  G. math and other functions  " link CH_9G }
          @{"  H. string and list linking functions  " link CH_9H }
          @{"  I. lisp-cells and cell functions  " link CH_9I }
          @{"  J. boopsi functions  " link CH_9J }

        10. library functions and modules
          @{"  A. built-in library calls  " link CH_10A }
          @{"  B. interfacing to the amiga system with the v47 modules  " link CH_10B }
          @{"  C. compiling own modules  " link CH_10C }
          @{"  D. the modulecache  " link CH_10D }

        11. quoted expressions
          @{"  A. quoting and scope  " link CH_11A }
          @{"  B. Eval()  " link CH_11B }
          @{"  C. built-in functions  " link CH_11C }

        12. floating point support
          @{"  A. float values  " link CH_12A }
          @{"  B. computing with floats  " link CH_12B }
          @{"  C. builtin float functions  " link CH_12C }
          @{"  D. float implementation issues  " link CH_12D }

        13. Exception handling
          @{"  A. defining exception handlers (TRY/CATCH or HANDLE/EXCEPT)  " link CH_13A }
          @{"  B. using the Raise() function  " link CH_13B }
          @{"  C. defining exceptions for built-in functions (RAISE/IF)  " link CH_13C }
          @{"  D. use of exception-ID's  " link CH_13D }

        14. OO programming
          @{"  A. OO features in E  " link CH_14A }
          @{"  B. object inheritance  " link CH_14B }
          @{"  C. data hiding (EXPORT/PRIVATE/PUBLIC)  " link CH_14C }
          @{"  D. methods and virtual methods  " link CH_14D }
          @{"  E. Constructors, Destructors and Super-Methods (NEW,END,SUPER)  " link CH_14E }
          @{"  F. Unions  " link CH_14F }

        15. inline assembly
          @{"  A. identifier sharing  " link CH_15A }
          @{"  B. the inline assembler compared to a macro assembler  " link CH_15B }
          @{"  C. ways using binary data (INCBIN/CHAR..)  " link CH_15C }
          @{"  D. OPT ASM  " link CH_15D }
          @{"  E. Inline asm and register variables  " link CH_15E }

        16. technical and implementation issues
          @{"  A. the OPT keyword  " link CH_16A }
          @{"  B. small/large model  " link CH_16B }
          @{"  C. stack organisation  " link CH_16C }
          @{"  D. hardcoded limits  " link CH_16D }
          @{"  E. error messages, warnings and the unreferenced check  " link CH_16E }
          @{"  F. compiler buffer organisation and allocation  " link CH_16F }
          @{"  G. register allocation  " link CH_16G }

        17. Essential E Utilities / Applications
          @{"  A. ShowModule  " link CH_17A }
          @{"  B. ShowHunk  " link CH_17B }
          @{"  C. Pragma2Module / Iconvert  " link CH_17C }
          @{"  D. ShowCache / FlushCache  " link CH_17D }
          @{"  E. ecompile.rexx  " link CH_17E }
          @{"  F. o2m  " link CH_17F }
          @{"  G. E-Yacc  " link CH_17G }
          @{"  H. SrcGen  " link CH_17H }
          @{"  I. EBuild  " link CH_17I }
          @{"  J. EE / Aprof  " link CH_17J }
          @{"  K. EDBG  " link CH_17K }
          @{"  L. E-VO PreProcessor  " link CH_17L }
          @{"  M. E-VO Library mode  " link CH_17M }

        18. Appendices
          @{"  A. E Grammar description  " link CH_18A }
          @{"  B. Tutorial  " link CH_18B }
          @{"  C. Mapping E to C/C++/Pascal/Ada/Lisp etc.  " link CH_18C }
          @{"  D. Amiga E FAQ  " link CH_18D }
          @{"  E. TODO/BUG list  " link CH_18E }




@endnode
@node CH_0A
@title "0A. introduction"
0A. introduction
----------------
E is an object oriented / procedural / unpure functional higher programming
language, mainly influenced by languages such as C++, Ada, Lisp etc.  It is
a general-purpose programming language, and the Amiga implementation is
specifically targeted at programming applications.  Features of the language
include:  speed of >20000 lines/minute on a vanilla A500, inline assembler
and linker integrated into compiler, large set of integrated functions,
great module concept with v47 (OS3.2) includes as modules, flexible
type-system, quoted expressions, immediate and typed lists, parametric and
inclusion polymorphism, exception handling, inheritance, data-hiding,
methods, multiple return values, default arguments, register allocation,
fast memory management, unification, LISP-Cells, gui-toolkit,
(macro-) preprocessor, very intuitive and powerful source-level debugger,
easy .library linking, and much more...

This is what 'HelloWorld' looks like in E:

/* nominated for Most Boring Example */

PROC main()
  WriteF('Hello, World!\\n')
ENDPROC


@endnode
@node CH_0B
@title "0B. the distribution"
0B. the distribution
--------------------
The distribution includes:

Bin/            contains the compiler E-VO and the support utilities
Modules/        directory containing all Amiga v47 E modules and
                useful tools as linkable modules
Docs/           documentation on E
Src/            example sources in E

This distribution of E-VO Amiga E is FreeWare, and may be freely copied.
It contains the compiler and modules, example code and various additional
tools originally distributed as part of the original Amiga E package which
were also released into the public domain.

Distributing copies of E-VO for more than the price of media+postage, or any
other way of distribution with only the slightest aim of making profit,
is not allowed.

No guarantees, no warranty.  If you manage to drown your pet goldfish
using E, or E fails to be suitable for ordering pizzas, that's entirely your
problem. Whatever happens, don't blame it on me.

@endnode

@node CH_0C
@title "0C. installing/using the compiler"
0C. installing/using the compiler
----------------------

To install E-VO on your system, just copy the whole distribution to some
place in your system, extend your path to the BIN directory, and assign
EMODULES:  to the 'Modules' directory.  If you're upgrading from a previous
version, please reinstall this release which includes the latest OS 3.2
modules and compiler.

syntax of the compiler (2.04+):

SOURCE/A,REG=NUMREGALLOC/N/K,LARGE/S,SYM=SYMBOLHUNK/S,NOWARN/S
QUIET/S,ASM/S,ERRLINE/S,ERRBYTE/S,SHOWBUF/S,ADDBUF/N/K,COMPILESTACK/N/K,
IC=IGNORECACHE/S,HOLD/S,WB/S,LINEDEBUG/S,OPTI/S,LEGACY/S,LEGACYINT/S,
DEBUG/S,DEBUG50/S,NILCHECK/S,STACKCHECK/S,ARRAYCHECK/S,SHOWFNAME/S,NOPROGRESS/S,DEFINE/M',0

for 1.2 to 2.03:

evo [-opts] <sourcefile>

As an example we'll compile the program 'HelloWorld.e'. The compiler will
produce an executable 'HelloWorld'.

1>  cd e:src
E:Src>  evo helloworld

E-VO E Compiler/Assembler/Linker/PP v3.4.0-rc1 (14.09.2021)
developing as "EC" '1991-97 by Wouter van Oortmerssen
developing as "GEC" '1999/2003 by grio
developing as "E-VO" '2021 by Darren Coles
lexical analysing ...
parsing and compiling ...
no errors
E:Src>  helloworld
Hello, World!
E:Src>  list
HelloWorld.e                  89 ----rwed Oggi      17:37:00
helloworld                   656 ----rwed Oggi      17:37:00
2 files - 4 blocks used
E:Src>

note: the compiler cannot be made resident

Options.
These are standard AmigaDos options. You can see them at any time by typing
'EVO ?'. For 1.2 options the old unix-like switches are used (need to be
written together, preceded by a "-"):

LARGE        -l  compiles with large code/data model (see @{" 16A " link CH_16A }, OPT LARGE)
ASM          -a  puts E-VO into assembler mode (see @{" 15D " link CH_15D }, OPT ASM).
NOWARN       -n  suppresses warnings (see @{" 16A " link CH_16A }, OPT NOWARN)
SYM          -s  adds a symbolhunk to the executable, for use with
                 profilers, debuggers, disassemblers etc.
LINEDEBUG    -L  adds a "LINE" debughunk to the executable, for use
                 with profilers, debuggers etc. (see @{" 17K " link CH_17K }).
REG=N        -rN uses N regs per PROC for register-allocation.
                 (default is 0 for the time being, read elsewhere
                 about how to use this!)
QUIET        -q  if there are no errors or warnings, E-VO won't output
                 anything.
WB           -w  puts wb to front (for scripts)
SHOWBUF      -b  shows buffer memory usage information
ADDBUF=X     -mX forces E-VO to allocate more memory for its buffers.
                 X ranges 1..9, the minimum number of 100k blocks to
                 allocate. default is 1 (never needed).
COMPILESTACK     Allows you to override the size of the compilestack in case
                 you get workspace full errors.
ERRLINE      -e  give the line# the error was on as returnvalue
ERRBYTE      -E  give the byte-offset in the file the error was on
                 as returnvalue
IGNORECACHE  -c  don't use the module-cache in any way
HOLD         -h  wait for a <return> on the commandline before
                 exiting E-VO.
OPTI             enable all optimisations (currently equals REG=5 and
                 STRMERGE on)
LEGACY           enable legacy mode - outputs module files in version 10
                 format and disable the new constants and built in functions
                 that may cause naming clashes with older code.
LEGACYINT        see OPT LEGACYINT. Enables backward compability with
                 Int() function
DEBUG            attach debug infos to executable/module (see @{" 17K " link CH_17K })
DEBUG50          attach debug infos to executable/module in a way that
                 allows debugging under OS 4.x (see @{" 17K " link CH_17K })
NILCHECK         insert code at every "." or "[]" dereference to check
                 for a NIL pointer. Will throw an exception of the form
                 Throw("NIL",source_line_number). Using this option will also
                 result in NIL checking being enforced in many of the E built
                 in functions which accept pointers.
ARRAYCHECK       insert code at every array access to check for array bounds. Will
                 throw an exception of the form Throw("INDX",source_line_number)
STACKCHECK       add code to the start each PROC to check for enough stack space.
                 Will throw an exception of the form
                 Throw("STCK",source_line_number)

SHOWFNAME        Displays the filename of the E file being compiled.
NOPROGRESS       Don't display progress line numbers during compilation
DEFINE           Equivalent to #DEFINE in your code, defines a macro

example:  evo large blabla
compiles blabla.e with large model.
NOTE: in most cases you won't need to use any of these options


@endnode
@node CH_0D
@title "0D. changes / history / new features"
0D. changes / history / new features
------------------------------------
 New in E-VO  (3.9.3)
   - fix address error on 68000 due to mis-aligned data causing crash on loading
   - fix code generation error when compiling eg arr[i].bb+=10
   - fix errors when using NEW obj.create() syntax
   - fix error with NOT operator bleeding into next field when using eg
       [0,NOT(0),1]
   - O2M: improved hunk handling

 New in E-VO  (3.9.2)
   - fix parsing issue with the : (colon) character could cause errors when
     parsing expressions using the ternery operator (?)
   - fix register corruption causing corruption of module files in some
     scenarios
   - fix casing of CharInstr (changed to CharInStr) and CharInstri (changed to CharInStri)
   - added EVO_3_9_2 constant
   - FindModule: Issue with freezing on some modules
   - ShowModule: Issue with freezing on some modules
   - Updated EVO Language Guide file to fix some errors / inconsistencies
   - Added unittests.e file and set up to run this for each build.
   
 New in E-VO  (3.9.1)
   - Fixed issue with TextF function crashing
   - Add cheeck for declaring exported variable that references
     a private object
   - Added support for PTR to imported object types in modules
   - Fixed issue with BYTE and WORD declarations followed by comma
   - Array of char/int/long etc declared in module caused crash
   - Array of object declared in module was not allocated the correct
     amount of stack space
   - Added NOPROGRESS command line switch to disable printing of progress
     line numbers during compilation
   - Updated to module format version 16
   - Added EVO_3_9_1 define
   - FindModule wasn't displaying PTR types correctly in older module
     versions
   - FindModule was showing signed / unsigned types reversed
   - FindModule: Add support for PTR to imported object types 
   - FindModule: updated to more correctly handle private objects and
     private globals
   - ShowModule wasn't displaying PTR types correctly in older module
     versions
   - ShowModule was showing signed / unsigned types reversed
   - ShowModule: Add support for PTR to imported object types 
   - ShowModule: updated to more correctly handle private objects and
     private globals
   - Added version string to FindModule
   - Fixed installer script incorrectly removing assigns
   - Updated module libraries/cybergraphics.m

 New in E-VO  (3.9.0)
   - FindModule: fix division by 0 error
   - FindModule: updated to handle module format version 15
   - EDBG: fix issues loading files generated with E-VO 3.8.x
   - O2M: Added support for symbol hunks
   - ShowModule: Updated to handle module format version 15
   - Warnings are now displayed when encountered rather than at the
     end of the compile
   - Add "assign of ARRAY/STRING/LIST variable" warning
   - Allow objects to be constructed using NEW objectname as well as
     NEW objectvar
   - When using OPT POOL, all memory allocation/free via NEW, String(),
     List(), New(), FastNew(), NewR(), END, Dispose(), FastDispose(),
     FastDisposeList() will be done from the pool
   - Added CharInStr(str,char) function to search for a char in a string
   - Added CharInStri(str,char) function, as CharInStr but case insensitive
   - Added MemCompare(mem1,mem2,len) function to compare two memory areas,
     returns -1 (mem1 contents less),1 (mem2 contents less) or 0
     (both equal)
   - Optimised WriteF, PrintF, PutF, TextF to skip the formatting part if no
     additional parameters are passed after the format string
   - Add NILCHECK variants of many E functions that raise a "NIL" exception
   - Small changes to ASTRCLONE and STRREM to remove unnecessary instructions
   - Exception raising code reworked to remove some code duplication
   - Global variables with default values now allowed in modules
   - Global array of object variables now allowed in modules
   - Added STACKCHECK command line option to add code to each PROC to check
     for enough stack space
   - Fix astrclone to always clear first byte of result if the length is 0
   - Fix for NILCHECK version of DoMethod which was not setting up d0
     correctly
   - Updated module format version to 15
   - Added STACKCHECK procedure modifier to enable stack checking for a
     particular procedure
   - Added GETA4 procedure modifier to restore A4 at the start of the
     procedure
   - Replaced all CLR <reg> to MOVEQ #x,<reg> or EOR as appropriate
   - add PROC x() MODSETUP and PROC x() MODCLEANUP - from creative but with
     name changed from SETUP/CLEANUP due to naming clashes
   - Fix memory corruption issue during parsing
   - fix error generating HUNK_DREL32SHORT hunk causing corrupt exe
   - fix legacy mode memory corruption creating corrupt module
   - increase fixed buffer sizes
   - fix assembler error with MULS.L and DIVSL.L commands generating
     incorrect opcodes
   - PTR indexes are now 32 bit so can be >65535 and <0
   - indexed ptr to object code generation optimised for 68020
     (to use MULS.L) and to generate add/shift/subtract combinations
     instead of MULS where appropriate
   - Fix for obj.prop+=2 which was broken due to register value not
     being preserved.
   - Add NIL test in Free()
   - disable 020 JSR to BSR.L optimisation when multiple sections in
     use (which was causing invalid code generation)
   - Add compile time array bounds checking (for constant array indexes)
   - Add ARRAYCHECK command line parameter to add code to each array access
     to check array bounds
   - Fix symbol hunks being generated incorrectly when multiple sections
     in use
   - added EVO_3_9_0 define
   - added newer version of cybergraphics module

 New in E-VO  (3.8.1)
   - FindModule: add module version 14 support
   - Showmodule: add module version 14 support
   - fix legacy compile mode so it doesnt include changes in module
     version 14
   - Fix reaction texteditor tag values in gadgets/texteditor.m
   
 New in E-VO  (3.8.0)
   - Added variable names + types and procedure line numbers to EVAR debug
     info
   - Made multi-dimensional arrays memory layout more logical (first
     dimension is most significant)
   - Fix global variable structure generated for library definitions
   - Added object definitions to debug info
   - Update EDBG to parse new EVAR structures
   - Add _SRCLINE_ define (current line number)
   - Add _SRCPROC_ define (current proc name)
   - Added EVO_3_8_0 define
   - Added support for ARRAY OF PTR TO ... in objects
   - Fixed issue where two variables declared with the same initial value
     did not get set correctly.
 
 New in E-VO  (3.7.0)
   - Compiler sometimes produced garbage error message with an error at the very
     end of the file.
   - Reformatted the Amigaguide document to better fit in 80 columns.
   - added Arexx compile scripts
   - allow creating a new object directly into an object member
     eg. self.a:=NEW self.a.method()
   - method table was not correctly initialised for objects declared as
     eg. DEF a:b OR DEF a[5]:ARRAY OF b
   - heap in compiler is allocated with MEMF_CLEAR in compiler to ensure all
     data in the exe is initialised to 0 and not contain random data from
     memory
   - calling a method on an array element of an object did did not calculate
     the correct address for the method.
   - compiler was creating 020 opcodes when accessing an array of objects
   - added NIL pointer checks when calling methods on array of objects
   - self variable was set incorrectly for methods declared with an exception
     handler
   - fix issue with wrong opcode being generated in this code
        DEF a
        DEF b:PTR TO LONG
        a:=[0,b[0]:=5]
     resulting in the value not being stored correctly
   - include textfield module  
   - fix bug in reaction_lib with the freeClickTabs, freeChooserLabels and
     freeRadioButtons functions not looping through the list correctly.
   - add support for long branch 020 instruction.
   - fix invalid memory access when casting the result of a procedure call
     to an object
   - fix inconsistencies with SIZEOF command
   - fix error in branching code for the == operator
   - add missing listview module
   - add some extra functions in reaction_lib (openClass, openLibs,
     closeLibs, libDoGadgetMethodA)
   - added PSIZEOF operator
   - added ARRAYSIZE operator
   - added SelectFirst function
   - added debug symbols for methods
   - allow CONT and EXIT within SELECT statement
   - updated Rnd() function to be pure
   - flagged openw, opwns, boolgad, settopaz, sets, gets functions as not pure
   - added WARN TRUE/FALSE to enable/disable warnings
   - added Rol() and Ror() functions
   - the 020 variant of Mod() was not correctly used by the compiler and
     was also taking the parameters the wrong way around
   - added 020 variant of ListItem function
   - track line numbers for warnings
   - NOREGS option was not correctly parsed on methods
   - taking a reference to a method {obj.method} did not return the
     correct address
   - 1.3 startup code now updates stack pointers to allow stack space
     functions to return the correct values.
   - skip NameFromFH call on kick 1.3 meaning the compiler now works
     on that OS again
   - NOT operator was very broken (inconsistently applied and in some cases
     not applied at all)
   - Added EVO_3_7_0 define
   - casting and PTR TO PTR handling is improved and should now work in all
     use cases.

 New in E-VO  (3.6.1)
    - add checks for object going over 64k
    - fix 020 optimised address calculation for array of object (or ptr to
      object) indexing
    - added EXENAME and DESTDIR command line arguments to override the
      output file name.
    - fix casing for IeeeDPFloor and IeeeDPCeil
    - Extend lists to handle 64k items instead of 32k.
    - Fix #else preprocessor parsing issues

 New in E-VO  (3.6.0)
    - EDBG: add support for ECX/EEC compiled executables
    - fix issue with memfill that could cause odd address exception on 68000
    - add TRY..CATCH..ENDTRY allowing more control over exception handling
    - Add UNTILN, WHILEN ELSEWHILEN, IFN, ELSEIFN, EXITN, CONTN which are
      inverse of their counterparts eg. UNTILN x is equal to UNTIL Not(x)
    - Add WORD and BYTE types (unsigned 16 bit and signed 8 bit variable
      types) and Word() Byte() functions as well as PutByte() and PutWord()
    - Allow SIZEOF object and SIZEOF object.member
    - Fix issue where SIZEOF x didnt throw a compile error if x was not
      defined
    - Added EVO_3_6_0 define
    - Added Compare, Ucompare, Fcompare, StrCompare functions
    - Fixed Int() function to sign extend result
    - Added OPT LEGACYINT to restore old version of Int() command
    - Added command line switch LEGACYINT/S which does the same as the
      above OPT
    - Allow BYTE and WORD types in modules, increase module version to 13.
    - FindModule: fix error in memory deallocation
    - FindModule: update for BYTE and WORD types (module version 13)
    - ShowModule: update for BYTE and WORD types (module version 13)
    
 New in E-VO  (3.5.1)
    - fix issue with short jump calculations being done incorrect in certain
      circumstances, causing crashes.
    - fix casing of math ieee functions to match original E modules
    - disable OPT POOL and OPT UTILLIB in legacy mode
    - disable THISTASK, __POOL and UTILITYBASE in legacy mode
    - reinstate [] in immediate list to previous E functionality (a blank
      0 item list)
    - fix optimisation issue where MOVE.B (A0),D0 and MOVE.W (A0),D0
      followed by TST.L would be optimised incorrectly.
    - fix issue with immediate lists, ListMax() could not be used as the max
      size was not populated.
    - allow 0 length lists to be created eg List(0)
	- add showhunk source
 New in E-VO  (3.5.0)
    - Added ac020 object with pad int to allow optimised 32 bit writes on
      020+ courtesy of Samuel D. Crow
    - When compiling in debug, disable the optimisation that removes LINK,
      UNLK in procs where not needed (EDBG doesnt step properly when these
      aren't present)
    - EDBG: Add case insensitive search
    - EDBG: Added search next
    - EDBG: Display variable values in both hex and decimal
    - Fixed: Clears an extra byte in the String() initialisation causing
      memory corruption in the case of String(0)
    - EDBG: Enhanced breakpoints - multiple can now be set and can be
      seen/set in a column next to the code
    - EDBG: Support for multiple breakpoints and debugging while code is
      running.
    - EDBG: Added support for debugging under OS4.x (eg without using CPU
      exceptions)
    - EDBG: Added support for adding watch variables by entering name
    - Added support for DEBUG50 mode which allows debugging using JSR
      instructions instead of causing cpu exceptions which allows debugging
      under OS 4.x
    - Added support for multi-dimensional arrays
    - Added support for PTR TO PTR variables
    - Added EVO_3_5_0 define
    - Added ? operator (ported from CreativE)
    - Added OPT FPEXP (fpu floating point support - ported from CreativE)
    - Added :=: swap operator (ported form CreativE)
    - Added OPT INLINE (ported from CreativE)
    - Added OPT NOSTARTUP (ported form CreativE)
    - Added OPT UTILLIB (ported from CreativE)
    - Added NOREGS option (ported from CreativE)
    - Updated module format (now version 12) to account for multi-dimension
      arrays and PTR TO PTR type in object members
    - Ported some optimisation code from CreativE
    - Added ENDPROC WITH function (ported from CreativE)
    - Added == operator (range check - ported from CreativE)
    - Added new system constants TAG_DONE, TAG_END, TAG_INORE, TAG_MORE,
      TAG_SKIP, TAG_USER, OFFSET_BEGINNING, OFFSET_BEGINING, OFFSET_CURRENT,
      OFFSET_END (ported from CreativE)
    - Added new utilitybase variable
    - Added 020+ optimised MOD function
    - Added support for += -= etc operators (ported from CreativE)
    - ShowModule: Added support for new module format (version 12)
    - Added improved LONG,INT,CHAR routines from CreativE which allow
      expressions
    - Added improved SIZEOF command that allows SIZEOF objectvar
    - Allow \xNN for hex char in strings
    - Add ELSEWHILE and ALWAYS extension to WHILE LOOP (from CreativE)
    - Improvements to SELECT x to allow expressions (from CreativE)
    - Added LEGACY option to output module format version 10 and disable
      newer features
    - Fix: 2 << obj.val was generating corrupt code
    - Add SECTION (code and data) support
    - Fix: ReadStr return value was not completely compatible with EC (fix
      Pragma2Module bug)
    - ShowModule: Add -e parameter which will force output into a form that
      can be directly recompiled
    - Added a scanf module in the other folder
    - Allow string consts in modules
    - Allow float CONST definitions
    - Remove some unused startup code when OPT ASM in use.
    - ShowModule: Added support for string CONSTs
    - Added OFFSETOF to get the offset of a member of an object
    - Added MemFill, AstrClone, ListInsItem, ListRemItem, ListSwapItem
      functions
    - EDBG: Support for sections
    - FindModule: Coded new FindModule application from scratch with support
      for latest module format.
    - Make second parameter of SetStr optional and if not specified it will
      set the calculate the string length automatically
    - Added EndsWith function eg EndsWith(s,'A') returns TRUE if a string1
      ends with string2
    - Fix: CHAR '' or CONST STR='' would cause memory corruption
    - Fix "3*SIZEOF LONG" error
    - added AstpCopy and UsedStack and StackSize functions
 New in E-VO  (3.4.1)
    - Fix: Optimisation error where multiple variables set to the same value
      were not set correctly.
    - Made the select source file window wider in EDBG to account for the
      full path being displayed.
 New in E-VO  (3.4.0)
    - Name change to E-VO
    - New version numbering 3.4.0
    - Add CONT command, similar to EXIT but continues the loop at the top
      instead of dropping out.
    - Make default cpu 68000
    - Make stack buffer configurable (and default it to 50k)
    - Fix file handle leak
    - Fixed breaking defect MIDSTR command
    - Allow CONT and EXIT commands to be called inside an IF or SELECT
    - Add ANDALSO and ORELSE short circuit versions of AND and OR
    - Optimised IF/WHILE/REPEAT expressions and removed TST immediately
      following MOVEQ
    - Fixed KickVersion not returning true when version number is exactly
      equal to the requested version
    - Allow String(0) which was valid in E v3.3a
    - Show line numbers for unused variables/labels/procs
    - Allow new formatting characters \u (unsigned long) \q (double quote)
      \! (ascii bell) \v (vertical tab) \xHH (hex char) (taken from
      CreativE)
    - Allow macros to be defined by passing in DEFINE command line
      parameter (taken from CreativE / ECX)
    - Allow string consts (taken from CreativE)
    - Added #date preprocessor (taken from CreativE but enhanced with hours
      minutes and seconds options)
    - Added DoMethod, DoMethodA, CoerceMethod, CoerceMethodA,
      DoSuperMethod, DoSuperMethodA, Set, Sets, Get, Gets, CtrlD, CtrlE,
      CtrlF, Chk, Eof, Fopen, Fclose, Alloc, Free, PutF, ReadB, WriteB,
      Size, Lsl, Lsr functions (all taken from CreativE)
    - We now have optimised versions of WriteF, PrintF, PutF when targetting
      OS 37.
    - Added SHOWFNAME command line parameter.
    - Reverted order of efunctab back to how it was in E to restore
      compatibility with pre-compiled E modules that are using Mul Div or
      WriteF.
    - Added the following functions:
        LowerChar(c), UpperChar(c) - takes a char and returns a char
        StrAddChar(s,c) - Adds a single character to an estring
        ListAddItem(l,i) - Adds a single item to an elist
    - Immediate assignments for values -256 to 254 for even numbers
      converted to MOVEQ followed by ADD.L D0,D0
    - Optimised NEW [] list creation to remove duplicate moves into D0
      eg. NEW [0,0,0,0]
    - Added OPT PURE to check for non pure code (eg variables in static
      lists/objects and GetA4 call)
    - Optimisation: Removed TST.L D0 after EXT.L D0 in IF/WHILE/REPEAT etc
      conditions.
    - Optimisation for default values for reg vars into single MOVE
      instruction.
    - Optimisation for reg var assignments where moveq is used into a single
      MOVE instruction where possible.
    - Added warning when compiling a library that OPT PURE is recommended
    - Optimisation for reg var assignments where moveq+add is used.
    - Optimisations with FOR loops to remove redundant move instructions.
    - Added FATAL 'string' command which generates no code but will fail
      the compilation (useful for use with #ifdef)
    - Added WARN 'string' command which generates no code but will show
    - a warning during compilation
    - Added GEC_3_4_0 auto generated #define macro which will allow you
      to check for a minimum version of the compiler. New defines will be
      additionally created for each new version of the software.
    - Added #ELSE #ELIFDEF #ELIFNDEF preprocessor commands
    - Added #UNDEF preprocessor command
    - Added 020 optimised versions of Long, Int and Char functions
    - Added StrIns(estring,estring,pos) - insert string function
    - Added StrRem(estring,pos,len) - remove substring function
    - Store full path of source files when generating debug hunk (allows
      EDBG to find the files if they are not all located in the same
      folder).
    - v47 (OS3.2) emodules, including all the new reaction interfaces
    - ARRAY OF CHAR no longer word aligned in OBJECTS
    - NOT operator works correctly with immediate values (eg "NOT 0" did not
      work correctly.
    - MUL/DIV optimisations also working for 020 generated code.
    - NOALIGN now requires 020 as accessing unaligned INT or LONG will
      result in exception if you don't have 020.
    - Fixed optimiser bug in library calls where IF function was used in
      a parameter that prevented the correct value from being passed.
     - Fix: x AND 0 was returning incorrect value
     
 New features:
 (3.3g)
    - Improved fake stack setting , now is more os compatible
      ( used exec SwapStack() for OS v37 else old things ).
    - Added PRIVATE methods in OOP ( PROC bla() OF abc PRIVATE IS EMPTY )
    - member can be used as method ( OBJECT abc;a,b;ENDOBJECT -> abc.a(1,2) )
    - possible set PROC/label address in static list like this [1,2,main]
      ( old mode [1,2,{main}]  ; is still working  ,but code is longer )
    - registers can be used as normal variable ( very beta )
    - EXTRA keyword for LIBRARY mode (idea taken from CreativE) :
      LIBRARY namestring,version,revision,versionstring EXTRA value IS ...;
      now library base can have more variables
    - PREPROCESS'ing is on at beginnig , OPT PREPROCESS leaved for down
      compatibility.
    - Added shift expresion (exp. CONST  ABC=1<<4; x:=y<<3 ).
    - Function added AstringF - like StringF but for normal string's
 
Fixed Bugs:
 (3.3g)
    - Another bugs fixed in "get address " sequence (ex.: {a.c[a]} ).
    - Problem with definition static OBJECT's and UNION's members.
    - Small bug in reg exp

New features:
 (3.3f)
    -New functions:
        StrClone(estr) - duplicate estring
        ListClone(elist) - only long size elements
        GetA4() - ...
    -Keyword SAFE for PROC's ( ex: PROC SAFE bla(a,b) ) :
        result push REG's D2-D7/A2-A6 on the stack .
    -OPT RUNBG - run background ( 2 vers 4 kick<v37 & kick>=v37 ) .
    -Modified startup code (to work with or without RUNBG , and
        fake stack is setup ).
    -New global var "thistask" .
    -Modified LIBRARY startup/Open/Close ; tasks never be hash-tabled
     but tabled by dynamic allocated one linked list .
    - "//" is a comment .
    -Added exporting in MODULE arg name for method's ;
         optional can be switched off by new keyword OLD
     ( ex: OPT MODULE OLD ); update ShowModule tool for display "new"
     module .

Fixed Bugs:
 (3.3f)
    - Bad reporting in case: OBJECT bla; a,b,c ;PROC zzz() ( or CONST ...)
      error message "label expected" - now "incorrect object definition" .
    - Small fixes in head code in LIBRARY output .
    - Raise() in LIBRARY if PROC isn't have HANDLE'r JUMP to $0 address .
    - Small bug in geting var address .
    - Agrrh... nasty bug in close library routine . Resolved.
    - Fake stack maker bug ( in setting Upper/Lower pointers) .

Optimisations:
 (3.3f)
    Some improvements in EAEXP.

New features:
 (3.3e)
    -Modified startup code .
    -REGS var in HANDLER PROC ( only if you define them ,
     no in optimisation ).Becarefull with that this maybe very dangerous
     if you use REGS var in HANDLE'r.
    -Using LONG label -> we get reloc of entered label .
    -Short version ARRAY definition in PROC/OBJECT ( DEF a[10]:CHAR ) .
    -NOALIGN keyword for OBJECT definition , it's turn off auto align 
     odd members (example : OBJECT blah NOALIGN ).
    -INC/DEC extension : INC x,20 ( x:=x+20 ) .
    -ASM mnemonics:
        MUL(U/S) AND DIV(U/S)   - long size
        EXTB.L
    -NOT keyword for IF/WHILE/UNTIL expresion .
    -UNION keyword for OBJECT's definitions :
           OBJECT abc
             a:LONG
             UNION
               [b:INT,c,d]
               [
                e:CHAR
                f:LONG
                g:INT
               ]
             ENDUNION
             k:PTR TO LONG
          -> or something like that :
             UNION [
              [b:INT,c,d],
              [e:CHAR,f:LONG,
               g:INT]
             ]
           ENDOBJECT
    -Added operators : &  eq AND
              || eq OR
              ~  eq NOT .
    -Possible assigment in member/list elements ([a.e:=0]) .

Fixed Bugs:
  (3.3e)
    -Getting address member if he not defined .
    -Bad brach set in LIBRARY if OPT CPU not used .
    -Things like:
        IF IF x THEN x ELSE x=y THEN y ELSE 0
     generating bad code .
    -"IF" optimiser cut TST after functions/procedures;
        Example:
        IF CtrlC() THEN Raise(ERR_BREAK)
        CODE:
        BSR     CtrlC
        BEQ     xxx .

New features:
 (3.3d)
    -In definition OBJECT or var array size can be defined by [4*3+1]
    -DEF a:REG abc now accepted .
    -"=>" and ">=" it's the same , "=<" and "<=" it's the same .
    -Things like that : Func()::blah.a now is possible .
    -Using OPT 020+ we get improved code :
     1. to the startup compiler put test code for entered CPU
     2. inline mul/div (*;/) working for long sizes
     3. function Mul()/Div() is patched for 020+
     4. EXT.W;EXT.L is changed to EXTB.L .

Fixed Bugs:
 (3.3d)
    -Get address member then a first is dereference : {a::abc.c} .

Optimisations:
 (3.3d)
    Removed LINK A5,#-XXXX/UNLK A5:
        if PROC haven't args and vars


    Put element to the list:

        Was:                    Is:
        LEA     xx(PC),A0       PEA     xx(PC)
        MOVE.L  A0,D0
        MOVE.L  D0,-(A7)

        Was:                    Is:
        MOVE.L  #xx,D0          PEA     xx.(W/L)
        MOVE.L  D0,-(A7)

        Was:                    Is:
        MOVE.L  xx(A4/A5)/Dx,D0 MOVE.L  xx(A4/A5)/Dx,-(A7)
        MOVE.L  D0,-(A7)


New features:
 (3.3c)
    -String joining in the list by "+" . 
    -Get/set list elements by 1+a .
    -Function InStri() -> like InStr() but it case sensitive .
    -Geting address of object member/method: {abc.a[3].z} .
    -Line number report in branches .
    -OPT STRMERGE for string merging .
    -EXPORT ignored if not used in MODULE .

Fixed Bugs:
 (3.3c)
    -Bad object reference:
        OBJECT bla
        brr
        ENDOBJECT
        PROC main()
        DEF i:bla
        WriteF('\d\n',i.brr.brr)
        ENDPROC
    -MOVEQ #-128,D0 now accepted .
    -Default member in object now is PTR TO CHAR .
    -RAISE error report .
    -Error report without "]" in DEF a[ .
    -Library base showed unref if called in other module .
    -EOR.L #3,D0 changed to EORI.L #3,D0 .

Optimisations:
  (3.3c)
    Short instruction for interval (-128;127):
        Was:                    Is:
            MOVE.L  xxxx,D0
        CMP.L   #0,D0           BEQ     xxx
        or
        TST.L   D0
        BEQ     xxx
        Was:                    Is:
        CMP.L   #1,D0           MOVEQ   #1,D1
        BEQ     xxx             CMP.L   D1,D0
                    BEQ     xxx

        Was:                    Is:
        OR.L    #-1,D0          MOVEQ   #-1,D1
                    OR.L    D1,D0

        Was:                    Is:
        SUB.L   #25,D0          MOVEQ   #25,D1
                    SUB.L   D1,D0

        Was:                    Is:
        ...                     ...

    MOVEM instrunction:
        Was:                    Is:
        MOVEM.L D3,-(A7)        MOVE.L  D3,-(A7)
        MOVEM.L (A7)+,A5        MOVEA.L (A7)+,A5

    Start value for variable:
        Was:                    Is:
        MOVEQ   #0,D0           MOVEQ   #0,D0
        MOVE.L  D0,-$0224(A4)   MOVE.L  D0,-$0224(A4)
        MOVEQ   #0,D0           MOVE.L  D0,-$0228(A4)
        MOVE.L  D0,-$0228(A4)

    NEW [1,x,10000]:
        Was:                    Is:
        MOVE.L  D0,(A7)         MOVE.L  D0,(A7)
        MOVEQ   #1,D0           MOVEA.L D0,A6
        MOVE.L  (A7),A6         MOVEQ  #1,D0
        MOVE.L  D0,4(A6)        MOVE.L  D0,4(A6)
        MOVE.L  x(A5),D0        MOVE.L  x(A5),D0
        MOVE.L  (A7),A6         MOVE.L  D0,8(A6)
        MOVE.L  D0,8(A6)        MOVE.L  #10000,D0
        MOVE.L  #10000,D0       MOVE.L  D0,12(A6)
        MOVE.L  (A7),A6
        MOVE.L  D0,12(A6)

    Zero offset to Ax register:
        Was:                    Is:
        MOVE.L  (A7)+,$0000(A0) MOVE.L  (A7)+,(A0)
        MOVE.L  D0,$0000(A7)    MOVE.L  D0,(A7)


New features:
  (3.3b)
    -Possible to use EXIT in LOOP/ENDLOOP and REPEAT/UNTIL .
    -Improved functions:
        StrCopy()
        StrCmp()
        StrAdd()
        MidStr()
        ForAll()
        Exitst()
        SelectList()
        FileLength()
        ReadStr()       ->      much faster
        MapList()
        ListCmp()
        ListAdd()
        ListCopy()
        String()
        OstrCmp() .
    -New function StriCmp() -> case sensitive version of StrCmp() .

Fixed Bugs:
  (3.3b)
    -Functions:
        Facos()
        Fsinh()
        ... . 

Optimisations:
  (3.3b)
    Better ELSEIF:
        Was:                    Is:
        Scc     D0              Bcc     xxx
        EXT.W   D0
        EXT.L   D0
        BEQ     xxx
    
    Testing data in D0:
        Was:                    Is:
            MOVE.L  xx(Ax),D0
                OR
            MOVE.L  Dx,D0
        TST.L   D0              Bcc     xxx
        Bcc     xxx
    
    Stack back:
        Was:                    Is:
            JSR     xxxPROC/EFUNC
        LEA     4(A7),A7        ADDQ.L  #4,A7

New in v3.3a:
- significantly improved version of EasyGUI, including the SOURCE!
- new version of EDBG which a.o. now cooperates with Explorer
  (see @{" 17K " link CH_17K } for a full list of new features and bugs fixed).
  Now comes with source code as well!
- new version of EBuild
- various new examples, e.g. a texturemapper
- new versions of many modules and examples
- revised docs, e.g. a new "known-bugs" list.
- some new builtins, e.g. the Arc float functions.
- LIBRARY mode now has a PROC close() to match main()
- two (!) new class libraries now part of the distribution,
  Gregor's OO modules and Fabio's AFC.

Bugs fixed in v3.3a:
- EC now saves debug information for "self", and EDBG makes use of it.
- OPT RTD removed.
- ENUM now works with negative numbers
- module-cache flushing now definitely fool proof.
- and many smaller bugs...

New in v3.2i:
- make sure you read about the bug in mathieeesingbas.library
  (see @{" 12D " link CH_12D }) (!)
- You can explicitly assign registers in own libraries (e.g. 'myfun(A0)').
- 'find' in EDBG
- Iconvert supports PhxAss (with "-p")
- quite a few things in the distribution

Bugs fixed in v3.2i:
- methods with an exception handler could cause problems in the register
  allocator.
- various bugs in EDBG (scroller etc.)
- double declaration of object or label identifiers could go unnoticed
- EC would have problems flushing modules from different directories (now
  it only looks for the file-part and flushes it).
- problems could occur with method being called before their OBJECT
  declarations.
- and various smaller bugs...

New features in v3.2e:
- better EasyGUI (plugins)
- better EDBG (Arexx port and config saving)

Bugs fixed in v3.2e:
- procident.member gave weird errors
- preprocessor failed on 68000 in particular situations
- preprocessor didn't handle international characters
- '\\0' in strings would cut them off
- EC would Lock() with an illegal mode from other dirs
- $FFFF as default arg became $FFFF0000
- macro errors could cause garbage to be printed on screen
- /* */ in comments failed
- inheriting from yourself wasn't detected
- didn't accept 0(A3,a) or 0(A3,a.L) or lab(PC,a) etc. where a:REG
- AND.L D0,a where a:REG changed src-dest

New features in v3.2a:
- EC can now output .library files!!! (see @{" 17M " link CH_17M })
- new EDBG with a massive amount of new features! (see @{" 17K " link CH_17K })
- enhanced EasyGUI (have a look)
- NEW and END now work on subobjects as well (see @{" 4K " link CH_4K }).
- NILCHECK/S, will insert code to check all "." and "[]" dereferences for
  NIL pointers (see @{" 0C " link CH_0C })
- v40 emodules, with types!
- new modules (Src/Class/ScrollWin/ and others)
- new docs / beginners guide
- extended distribution with Jason's RKRM examples etc.

Bugs fixed in v3.2a:
- IGNORECACHE was broken
- silly bug in /2
- asm -xx(Ax,Dx.w) addressing mode assembled wrong
- a-- where a:PTR TO o where SIZEOF o>8 translated into ADDI ipv SUBI
- globals in modules used in inline asm were not exported / offsets
  adjusted
- exp*globvar in module not compiled ok
- macros of other module would be saved again in current module
- and quite a few smaller ones...
     
a version history:

vers    date       k    comment

v0.3    sep 91   5      first version that outputs executables
v1.0    may 92  20      builtins and library calls, usable for small tools
v2.0    okt 92  38      v37 modules usable, language pretty "complete"
v2.1a   jan 93  44      first public release
v2.1b   apr 93  45      bug fix, few new features. current version for >1
                        year
v3.0a   jul 94  82      enormous amount of new features. E gets a price tag.
v3.0b   sep 94  82      bug fix
v3.0e   nov 94  83      bug fix
v3.1a   dec 94  85      EDBG, preprocessor etc.
v3.1i   jun 95  83      new EDBG etc.
v3.2a   jul 95  85      .library mode
v3.2e   nov 95  85      new EDBG, EasyGUI, bugfixes.
v3.2i   nov 96  86      bugfixes, various
v3.3a   okt 97  86      bugfixes, new EDBG/EasyGUI/modules/classes/sources
                        etc.

v3.3b   20.11.1999 -> 12.12.1999
v3.3c   15.12.1999 -> 12.01.2000
v3.3d   21.01.2000 -> 02.03.2000
v3.3e   05.04.2000 -> 22.06.2000
v3.3f   03.01.2001 -> 15.10.2001
v3.3g   04.01.2002 -> 16.06.2003
v3.4.0  09.06.2021 -> 07.10.2021
v3.4.1  19.10.2021 -> 23.10.2021
v3.5.0  04.11.2021 -> 14.07.2022
v3.6.0  12.09.2022 -> 19.04.2023
v3.6.1  24.04.2023 -> 16.06.2023
v3.7.0  30.06.2023 -> 09.03.2024

@endnode
@node CH_0E
@title "0E. additional information"
0E. additional information
--------------------------

The original Amiga E Compiler was developed over the course of more than six
years, after the author's idea of a good programming language, and a quality
amiga specific compiler for it.  It started out as quite a small project and
evolved over the years into the large development environment it is now.  EC
(and now E-VO) was programmed (as you might have guessed) in 100% assembly,
using the AsmOne assembler.  All other support programs and tools (such as
EDBG) were written in E itself.

Special thanks go to the following people:

Jason Hulance - for 'The Beginners Guide', modules/examples, helping out
  with EasyGUI/EDBG, betatesting, his registration site, and practically
  being the coauthor of the Amiga E package.
Gregor Goldbach - for his registration site and general help.
Jörg Wach - for EPD and general support.
Rob Verver - for comments/inspiration.
Barry Wills - for being such a good betatester.

I also would like to thank the following people for various reasons
(no particular order):

Raymond Hoving, Erwin van Breemen, Michael Zucchi, James Cooper, Jens Gelhar,
Paolo Silvera, Sergio Ruocco, Jeroen Vermeulen, Jan van den Baard,
Norman Kraft, Urban Mueller, Charles McCreary, Olivier Anh, Lionel
Vintenat, Rob Nottage, Mirko Lalli, Jonas Tehler, Paul Nolan, Antonio J.
Gomez Glez, Fabio Rotondo and many more...

@endnode
@node CH_1A
@title "1A. tabs,lf etc."
1A. tabs,lf etc.
----------------
E-sources are pure ascii format files, with the linefeed <lf> and
semicolon ";" being the separators for two statements. Statements
that have particularly many arguments, may be spread over several lines
by ending a line with a comma (or any other lexical element that can
normally never occur at the end of a line), thus ignoring the following
<lf> (see @{" 5A " link CH_5A } for line endings). Any lexical element in a source code
file may be separated from another by any number of spaces, tabs etc.


@endnode
@node CH_1B
@title "1B. comments"
1B. comments
------------
comments may be placed anywhere in a source file where normally a space
would have been correct. They start with '/*' and end with '*/'
and may be nested infinitely. The same goes for one-line comments,
which start with a '->' and end at the first <lf>.

In addition single line comments using // are also supported which were not
originally part of the Amiga E language specification.

/* this, is a comment */
-> this one too.
// another comment


@endnode
@node CH_1C
@title "1C. identifiers and types"
1C. identifiers and types
-------------------------
identifiers are strings that the programmer uses to denote certain
objects, in most cases variables, or even keywords or function names
predefined by the compiler. An identifier may consist of:

- upper and lowercase characters
- "0" .. "9" (except as first character)
- "_" (the underscore)

All characters are significant, but the compiler just looks at the
first two to identify the type of identifier it is dealing with:

both uppercase:               - keyword like IF, PROC etc.
                              - constant, like MAX_LENGTH
                              - assembly mnemonic, like MOVE
first lowercase:              - identifier of variable/label/object etc.
first upper and second lower: - E system function like: WriteF()
                              - library call: OpenWindow()

Note that all identifiers obey this syntax, for example:
WBenchToFront() becomes WbenchToFront()



@endnode
@node CH_2A
@title "2A. decimal (1)"
2A. decimal (1)
---------------
Immediate values in E all evaluate to a 32bit result; the only
difference among these values (A-G) is either their internal
representation, or the fact that they return a pointer rather than
a value.

A decimal value is a sequence of characters "0" .. "9", possibly
preceded by a minus "-" sign to denote negative numbers.
examples: 1, 100, -12, 1024

@endnode
@node CH_2B
@title "2B. hexadecimal ($1)"
2B. hexadecimal ($1)
--------------------
A hexadecimal value uses the additional characters "A" .. "F" (or
"a" .. "f") and is preceded by a "$" character.
Examples: $FC, $DFF180, -$ABCD

@endnode
@node CH_2C
@title "2C. binary (%1)"
2C. binary (%1)
---------------
Binary numbers start with a "%" character and use only "1" and "0"
to form a value.
Examples: %111, %1010100001, -%10101

@endnode
@node CH_2D
@title "2D. float (1.0)"
2D. float (1.0)
---------------
Floats differ only from normal decimal numbers in that they have
a "." to separate their two parts. Either one may be omitted,
not both. Note that floats have a different internal 32bit (IEEE)
representation (see @{" 12A " link CH_12A } for more information on floats).
Examples: 3.14159, .1 (=0.1), 1. (=1.0)

@endnode
@node CH_2E
@title "2E. character"
2E. character
-------------------
The value of a character (enclosed in double "" quotes) is its
ascii value, i.e. "A" = 65. In E, character immediate values
may be a short string up to 4 characters, for example "FORM",
where the first character "F" will be the MSB of the 32bit
representation, and "M" the LSB (least significant byte).

@endnode
@node CH_2F
@title "2F. string ('bla')"
2F. string ('bla')
------------------
Strings are any ascii representation, enclosed in single '' quotes.
The value of such a string is a pointer to the first character of it.
More specific: 'bla' yields a 32bit pointer to a memory area where
we find the bytes "b", "l" and "a". ALL strings in E are terminated
by a zero byte.
Strings may contain format signs introduced by a slash "\\", either
to introduce characters to the string that are for some reason
not displayable, or for use with string formatting functions
like WriteF(), TextF() and StringF(), or kick2 Vprintf().

\\n             a linefeed (ascii 10)
\\a or ''       an apostrophe ' (the one used for enclosing the string)
\\q             a doublequote: "
\\e             escape (ascii 27)
\\t             tab (ascii 9)
\\\\            a backslash
\\0             a zero byte. Of rare use, as ALL strings are 0-terminated
\\b             a carriage return (ascii 13)
\\xHH   print a character with the hex code HH
\\!             bell character (ascii 7)

Additionally, when used with formatting functions:

\\d     print a decimal number
\\u     print an unsigned decimal number
\\h     print a hexadecimal
\\s     print a string
\\c     print a character
\\z     set fill byte to '0' character
\\l     format to left of field
\\r     format to right of field (these last two act as toggles)

Field specifiers may follow the \\d,\\h and \\s codes:

[x]     specify exact field width x
(x,y)   specify minimum x and maximum y (strings only)

Example: print a hexadecimal number with 8 positions and leading zeroes:
WriteF('\\z\\h[8]\\n',num)

A string may be extended over several lines by trailing them with a "+"
sign and a <lf>:

'this specifically long string ' +
'is separated over two lines'


@endnode
@node CH_2G
@title "2G. lists ([1,2,3]) and typed lists"
2G. lists ([1,2,3]) and typed lists
-----------------------------------
An immediate list is the constant counterpart of the LIST datatype,
just as a 'string' is the constant counterpart for the STRING or
ARRAY OF CHAR datatype. Example:

[3,2,1,4]

is an expression that has as value a PTR to an already initialised list,
a list as a representation in memory is compatible with an ARRAY OF LONG,
with some extra length information at a negative offset. You may use these
immediate lists anywhere a function expects a PTR to an array of
32bits values, or a list. Examples:

['string',1.0,2.1]
[WA_FLAGS,1,WA_IDCMP,$200,WA_WIDTH,120,WA_HEIGHT,150,TAG_DONE]

(see @{" 9C " link CH_9C } on list-functions for a discussion on typed-immediate
lists and detailed information).


@endnode
@node CH_2H
@title "2H. lisp-cells (<a|b>)"
2H. lisp-cells (<a|b>)
----------------------
(see @{" 9I " link CH_9I } to read about cells in detail)



@endnode
@node CH_3A
@title "3A. format"
3A. format
----------
An expression is a piece of code held together by operators, functions
and parentheses to form a value.
They mostly consist of:

- immediate values (see @{" 2A " link CH_2A },...)
- operators (see @{" 4A " link CH_4A },...)
- function calls (see @{" 3D " link CH_3D })
- parentheses () (see @{" 3B " link CH_3B })
- variables or variable-expressions (see @{" 3C " link CH_3C })

examples of expressions:
1
'hello'
$ABCD+(2*6)+Abs(a)
(a<1) OR (b>=100)

@endnode
@node CH_3B
@title "3B. precedence and grouping"
3B. precedence and grouping
---------------------------
The E language has _no_ precedence whatsoever. This means that
expressions are evaluated left to right. You may change
precedence by parenthesizing some (sub-)expression:
1+2*3 /* =9 */     1+(2*3) /* =7 */       2*3+1  /* =7 */

@endnode
@node CH_3C
@title "3C. types of expressions"
3C. types of expressions
------------------------
There are three types of expressions that may be used for different
purposes;

- <var>, consisting of just a variable
- <varexp>, consisting of a variable, possibly with unary operators
  with it, like "++" (increment) "." (member selection) or [] (array
  operator). (see @{" 4D " link CH_4D }, 4G). It denotes a modifiable expression, like
  Lvalues in C.
  Note that those (unary) operators are not part of any precedence.
- <exp>. This includes <var> and <varexp>, and any other expression.

@endnode
@node CH_3D
@title "3D. function calls"
3D. function calls
------------------
A function call is a temporary suspension of the current code for a
jump to a function, this may be either a self-written function (PROC),
or a function provided by the system. The format of a function call
is the name of the function, followed by two parentheses () enclosing
zero to unlimited arguments, separated by commas ",". Note that arguments
to functions are again expressions. (see @{" 6A " link CH_6A }) on how to make your own
functions, (see @{" 9A " link CH_9A } and 10A)
on built-in functions. Examples:

foo(1,2)
Gadget(buffer,glist,2,0,40,80+offset,100,'Cancel')
Close(handle)



@endnode
@node CH_4A
@title "4A. math (+ - * /)"
4A. math (+ - * /)
------------------
These infix operators combine an expression with another value
to yield a new value. Examples:

1+2, MAX-1*5

(see @{" 12A " link CH_12A } on how to overload these operators for use with floats).
"-" may be used as the first part of an expression, with an implied 0,
i.e.    -a    or    -b+1      is legal.
Note that * and / are by default 16bit operators: (see @{" 9G " link CH_9G }, Mul())

If you have selected 020 cpu (OPT 020) then these will be swapped out with
020+ optimised versions which can handle 32 bit operations.

@endnode
@node CH_4B
@title "4B. comparison (= <> > < >= <=)"
4B. comparison (= <> > < >= <=)
-------------------------------
Equal to math operators, with the difference that they result in either
TRUE (32bit value -1), or FALSE. These can also be overloaded for floats.

@endnode
@node CH_4C
@title "4C. logical and bitwise (AND OR NOT << and >>)"
4C. logical and bitwise (AND OR << and >>)
--------------------------------
These operators either combine truth values to new ones, or perform
bitwise AND and OR operations. Examples:
(a>1) AND ((b=2) OR (c>=3))         /* logical */
a:=b AND $FF                        /* bitwise */

a:=NOT b (bitwise invert)

a:=1 << 3  /* bitwise shift left */
a:=128 >> 3  /* bitwise shift right */

@endnode
@node CH_4D
@title "4D. unary (SIZEOF PSIZEOF ARRAYSIZE OFFSETOF ^ {} ++ -- `)"
4D. unary (SIZEOF PSIZEOF ARRAYSIZE OFFSETOF ^ {} ++ -- `)
---------------------------------
- SIZEOF <objectident> or SIZEOF <var>
  simply returns the size of a certain object or CHAR/INT/LONG.  
  Example: SIZEOF newscreen + SIZEOF INT

  doing a SIZEOF on a pointer type always returns the size of the type
  pointed to. eg DEF a:PTR TO INT - SIZEOF a will be 2. One caveat
  to this rule is the default type. So although DEF a by itself will act
  as a PTR TO CHAR, in this case SIZEOF a will be 4 as it has not been
  specifically declared as a pointer type. Array types will also return
  the size of the individual elements not the whole size of the array.
  
- PSIZEOF <objectident> or PSIZEOF <var>
  works the same way as SIZEOF but all pointer types will return 4 (the
  size of the pointer. eg DEF a:PTR TO INT - SIZEOF a will be 4. Array
  types will still return the size of the individual elements.
  
- ARRAYSIZE {<dimension>,}<var>
  returns the number of elements declared in the array. The variable must
  be declared as an array for this to work. You cannot make a pointer
  variable and point it at an array and ask for an ARRAYSIZE of that pointer
  because the compiler calculates the size at compile time. You can also
  include dimension value, so if you want to get the size of a 2 dimensional
  array you must do ARRAYSIZE 1,var and ARRAYSIZE 2,var to get the 2
  dimensions seperately. eg DEF a[5][7]:ARRAY OF CHAR - ARRAYSIZE 1,a is 5
  and ARRAYSIZE 2,a is 7. The dimension cannot be a variable or calculation
  because as before the value is calculated at compile time.
  

- OFFSETOF <member>
  returns the offset from the base address of the object for that member
- {<var>}
  Returns the address of a variable or label. This is the operator
  you would use to give a variable as argument to a function by
  reference, instead of by value, which is default in E. See "^".
  [a bit obsolete for variables. use multiple returnvalues instead]
- ^<var>
  [a bit obsolete. you should be using the [] operator or multiple
   returnvalues instead]
  The counterpart of {}, writes or reads variables that were given by
  reference, examples:      ^a:=1     b:=^a
  it may also be used to plainly "peek" or "poke" LONG values from
  memory, if <var> is pointer to such a value.
  Example for {} and ^: write your own assignment function;

  PROC set(var,exp)
    ^var:=exp
  ENDPROC

  and call it with:     set({a},1)     /* equals a:=1 */
- <varexp>++   and  <varexp>--
  Increases (++) or decreases (--) the pointer that is denoted by
  <varexp> by the size of the data it points to. This has the effect
  that that pointer points to the next or previous item. When used
  on variables that are not pointers, these will simply be changed
  by one. Note that ++ always takes effect _after_ the calculation
  of <varexp>, -- always _before_. Examples:

  a++          /* return value of a, then increase by one */
  sp[]--       /* decrease pointer sp by 4 (if it were an array of long),
                  and read value pointed at by sp */

  Note that this is quite different from C, also, expressions like
  p.i++ work on p, not i.
- `<exp>
  This is called a quoted expression, from LISP. <exp> is not evaluated,
  but instead returns the address of the expression, which can later be
  evaluated when required. More on this special feature in chapter 11

@endnode
@node CH_4E
@title "4E. triple (IF THEN ELSE)"
4E. triple (IF THEN ELSE)
-------------------------
The IF operator has quite the same function as the IF statement, only
it selects between two expressions instead of two statements or blocks
of statements.

IF <boolexp> THEN <exp1> ELSE <exp2>

returns exp1 or exp2, according to boolexp. For example, instead of:

IF a<1 THEN b:=2 ELSE b:=3
IF x=3 THEN WriteF('x is 3\\n') ELSE WriteF('x is something else\\n')

write:

b:=IF a<1 THEN 2 ELSE 3
WriteF(IF x=3 THEN 'x is 3\\n' ELSE 'x is something else\\n')

You can also use the (less readable but more concise) C style syntax - x?y:z
eg.

b:= a<1 ? 2 : 3

The bitwise AND, OR can be used in IF THEN ELSE constructs and will work as
you might expect

IF (a<1) OR (a>10) THEN b:=2 ELSE b:=3

but because they are simply operating on a bitwise level there is no short
circuit evalaution. So you cannot do this for example:

IF (obj<>NIL) AND (obj.myval=3) THEN b:=2 ELSE b:=3

because both sides of the AND operator will always be executed and this will
result in an invalid memory access when obj.myval is referenced
if obj = NIL.

To address this E-VO also has ANDALSO and ORELSE constructs which perform
the same operation as AND and OR but will not perform the second part of the
test if the outcome is already determined by the first part. eg x AND y will
always be false if x is false and x OR y will always be true if x is not
false.

They are only designed for use in IF statements although you can use them in
a bitwise way the results are not defined. It is best to stick to just using
them with boolean expressions and expecting them to return FALSE or TRUE.

IF (obj<>NIL) ANDALSO (obj.myval=3) THEN b:=2 ELSE b:=3

This is perfectly acceptable with ANDALSO since the (abj.myval=3) won't be
executed if (obj<>NIL) is false, so there will be no invalid memory
referencing.



@endnode
@node CH_4F
@title "4F. structure (.)"
4F. structure (.)
-----------------
<ptr2object>.<memberofobject> builds a <varexp>
The pointer has to be declared as PTR TO <object> or ARRAY OF <object>
(see @{" 8D " link CH_8D } for these), and the member has to be a legal
object identifier. Note that reading a subobject in an object this
way results in a pointer to that object. Examples:

thistask.userdata:=1
rast:=myscreen.rastport

If the member you select is of type PTR TO <type>, you may use "."
and "[]" to dereference further values. If you select an ARRAY or
substructure in an OBJECT, the result is again a PTR.

@endnode
@node CH_4G
@title "4G. array ([])"
4G. array ([])
--------------
<var>[<indexexp>] (is a <varexp>)
This operator reads the value from the array <var> points to, with
index <indexexp>. The index may be just about any expression.
Note1: "[]" is a shortcut for "[0]"
Note2: with an array of n elements, the index is  0 .. n-1
Examples:

a[1]:=10           /* sets second element to 10 */
x:=table[y*4+1]    /* reads from array */
x.y[3]             /* accesses array in an object */

@endnode
@node CH_4H
@title "4H. float operator (!)"
4H. float operator (!)
----------------------
<exp>!<exp>
Converts expressions from integer to float and back, and
overloads operators + - * / = <> < > <= >= with float equivalents.
(see @{" 12A " link CH_12A } to read all about floats and this operator).

@endnode
@node CH_4I
@title "4I. assignments expressions (:=)"
4I. assignments expressions (:=)
--------------------------------
Assignments (setting a variable to a value) exist as statement
and as expression. The only difference is that the statement
version has the form <varexp>:=<exp> and the expression <var>:=<exp>.
The latter has the value of <exp> as result.
Note that as <var>:= takes on an expression, you will often need
parentheses to force correct interpretation, like:

IF mem:=New(100)=NIL THEN error()

is interpreted like:

IF mem:=(New(100)=NIL) THEN error()

which is not what you mean: mem should be a pointer, not a boolean.
but you should write:

IF (mem:=New(100))=NIL THEN error()

it's a good habit to parenthesize any assignment expression that is
part of another one, if not already disambiguated by other constructs
such as "bla(a:=1)", "b:=a:=1" etc.

@endnode
@node CH_4J
@title "4J. sequencing (BUT)"
4J. sequencing (BUT)
--------------------
The sequencing operator "BUT" allows two expressions to be written
in a construction that allows for only one. Often in writing
complex expressions/function calls, it would be nice to do a second
thing on the spot, like an assignment. Syntax:

<exp1> BUT <exp1>

this says: evaluate exp1, but return the value of exp2.
Example:

myfunc((x:=2) BUT x*x)

assign 2 to x and then calls myfunc with x*x. The () around the
assignment are again needed to prevent the := operator from taking
(2 BUT x*x) as an expression.

@endnode
@node CH_4K
@title "4K. dynamic memory allocation (NEW)"
4K. dynamic memory allocation (NEW)
-----------------------------------
the NEW operator is a powerful operator for dynamic memory allocation.
it has various forms:

(assuming DEF p:PTR TO whateverobj, q:PTR TO INT)

NEW p

is an expression that will allocate zero-ised memory for the object-size
p points to. the resulting pointer will be put in p, and is also
the value of the expression. if NEW fails to get memory, it raises
a "MEM" exception. 'NEW p' is therefore roughly equivalent with:

IF (p:=New(SIZEOF whateverobj))=NIL THEN Raise("MEM")

with the difference that p never gets any value if an exception
is raised, and that the former is of course an expression, not
a statement

but there's more: it's also possible to allocate an array dynamically:

NEW p[10]       /* array of 10 objects */

NEW q[a+1]      /* array of INT, size is runtime computed */

(this doesn't work when instantiating classes)
A problem with list [1,2,3] expressions sometimes is that they are static,
and when using these to build large datastructures they would need to be
created at run-time.  The dynamic equivalent to static lists may then
be used:

p:=[1,2,3]:whateverobj           /* static structure */

p:=NEW [1,2,3]:whateverobj       /* dynamicly allocated! */

this works with both lists and typed-lists, and also with arrays:

NEW [1,2,3]        /* constant-list, dyn. (note: not a like a listvar!) */
NEW [1,2,3]:obj    /* object */
NEW [1,2,3]:INT    /* array of INTs */

freeing memory allocated with any variations of NEW is done automatically
at the end of the program, or by hand with END or FastDispose().
(note: the only exception is NEW <list>, use FastDisposeList())

If you use NEW [...]:obj, and the number of fields is less than the one
present in the object, additional 0/NIL/"\\0" whatever fields will be added.
this allows extending objects without getting into trouble with allocations
like these.  (note that this is different from the static [...]:obj)

when you use NEW as a statement, multiple pointers may follow, i.e.:

        NEW a,b.create(),c

is valid.

special forms of NEW and END allow to use a PTR within an object, so you can
write things like

        NEW a.b.create()
        NEW a.b[50]
        END a.b

in the examples, the pointer 'a' is not affected, the pointer 'b' in the 'a'
object is used.

It is also possible to use NEW to create objects without having to declare
a variable of that specific type by using an object after the NEW command

eg.

OBJECT obj
  x:INT
  y:INT
ENDOBJECT

PROC create() OF obj
  self.x:=0
  self.y:=0
ENDPROC

a:= NEW obj

or

a:= NEW obj.create()

Using this syntax the variable "a" does not need to be declared of the specific
object type. If you wish to call a method or access a member then it would need
to be assigned to a variable of the correct type first.

This can be useful if you have several child classes and you may wish to create
an instance of any one of those child classes but store it in the same variable.


NEW allocated lists are not to be used with the Link() functions (see @{" 9H " link CH_9H }).

(see @{" 5M " link CH_5M }, END)
(see @{" 9F " link CH_9F } for the fast memory allocation functions NEW makes use of)

@endnode
@node CH_4L
@title "4L. unification (<=>)"
4L. unification (<=>)
---------------------
Unification allows a totally different style of programming
that will be familiar to you if you're used to either Logic
(ProLog), equational or functional (Miranda/Gofer/Haskell)
programming. Most of us when using structures/arrays whatever
are used to get values from it by selection ("." and "[]"),
in these languages however pattern matching is used.

in E:

exp <=> uni_exp

exp can be any expression, but in v3 it's only really useful
if it somehow is a pointer to a list. uni_exp is the pattern
that is used to match exp. All constants in uni_exp much
match to values in exp, variables are set to their respective
values. if something doesn't match, no variables get a value,
and the expression has the result FALSE. otherwise TRUE.
example:

a:=[1,2,3]
...
IF a <=> [1,x,y] THEN ...

this will succeed with x=2 and y=3. If list a were to be another
lenght than 3, the match would fail too. The fact that a is a list
in the first place is something you need to assure by yourself, E-VO
simply tries to fit the uni_exp on whatever value it gets as exp.
examples of FALSE:

a <=> [1,x]             -> wrong list-len
a <=> [1,4,x]           -> 4=2 fails
'bla' <=> [1,2]         -> unpredictable result / crash ?

The fun thing with unification is that you can do very complex matches,
and that if you take the first field or so as a constant telling what the
structure is, you have a nice form of dynamic typing. And, all the time
without using PTRs!

a slightly nicer example:

[BLA,[1,'burp'],['bla',"bla"]] <=> [BLA,[1,x],y]

binds:

x='burp', y=['bla',"bla"]

or even:

IF myexp <=> [PLUS,[MUL,a,1],[SUBS,[PLUS,c,d],e]] THEN RETURN a+c+d-e

maybe a silly example, but one could imagine doing complex stuff
like this in a compiler's code-optimizer. it equals this traditional code:

IF ListLen(myexp)=3
  IF myexp[]=PLUS
    IF ListLen(dummy:=myexp[1])=3
      IF (dummy[]=MUL) AND (dummy[2]=1)
        a:=dummy[1]
        IF ListLen(dummy:=myexp[2])=3
          IF dummy[]=SUBS
            e:=dummy[2]
            IF ListLen(dummy2:=dummy[1])=3
              IF dummy2[]=PLUS
                c:=dummy2[1]
                d:=dummy2[2]
                RETURN a+c+d-e
              ENDIF
            ENDIF
          ENDIF
        ENDIF
      ENDIF
    ENDIF
  ENDIF
ENDIF

only then a bit more optimal. As you see there's a lot of expressive
power involved in unification as compared to traditional selection-based
programming.

For now, the only thing allowed in unification are untyped lists,
(integer) constants, variables and LISP-Cells (see @{" 9I " link CH_9I } for that).
The future will see this expanded with strings, typed-lists/objects,
and even expressions [!]


@endnode
@node CH_4M
@title "4M. pointer typing (::)"
4M. pointer typing (::)
-----------------------
when you write an expression like 'p' where p is a PTR TO object
(and only objects), this allows you to dereference it as such.
The pointer typing operator allows you to change such a type on the fly:

DEF p:PTR TO mp                 -> message port

p.sigbit                        -> access it
p::ln.name                      -> access pointer as if it were a node

this is very useful for pointers that can point to different sorts
of things or objects that are part of each other, where of course
only one declaration is possible.

A second handy use is in OBJECTs that have members declared as LONG,
which are actually pointers. You may type it on the fly to dereference
it anyway:

mywindow.userport::mp.sigbit

here the window OBJECT has userport defined as LONG, so you wouldn't
be able to dereference it normally.

@endnode
@node CH_4N
@title "4N. swap (:=:)"
4N. swap (:=:)
--------------
using the swap operator you can exchange the contents of two variables
without needing to use an interim variable.

DEF a,b

a:=1
b:=2

a:=:b

will swap the contents of a and b.

The results will be stored in the last-used variables (i.e.
in a+b/c*d, "d" is the last one). Multiple use of it is possible
but the variables are set after each two (i.e. when You write sth like:

    a:=:b:=:c

  first a is swapped with b and then b is swapped with c)

IMPORTANT NOTE!

This "thing" is only for variables, NOT for objects and members. SO! If you
do sth. like:

  a.b.c:=:d.e.f

then ONLY POINTERS are set with expression values! In this case it is equal
to sth. like:

  t:=a
  a:=d.e.f
  d:=t.b.c

@endnode
@node CH_4O
@title "4O. quick compare (==)"
4O. quick compare (==)
--------------------

This sequence lets You compare a single expression against any bounds or
values [or even expressions, if needed] in a very fast way. The main use
is:

   <exp> == [<exp>, <lower> TO <upper>,...]

What is a good thing about it? Well, I think it's the speed. The output
code is much shorter and the check procedure exits as soon as one comparison
is true [for example, if an array has 100 entries and the first entry
matches the given expression, it exits immediately with TRUE]. It might be
very useful together with IF sentence.

    Example use:
    PROC checkrange(value)
        IF value*16-2 == [2, 10 TO 20, xyz()-blah+3] THEN
            DisplayBeep(NIL)
    ENDPROC

    is equivalent to:

    PROC checkrange(value)
        IF (value*16-2=2) OR ((value*16-2>10) AND (value*16-2<20)) OR
           (value*16-2=(xyz()-blah+3)) THEN DisplayBeep(NIL)
    ENDPROC

@endnode
@node CH_4P
@title "4P. increment decrement etc (+= -= etc)"
4P. increment decrement etc (+= -= etc)
---------------------------------------
That works mostly same as in C. You can easily omit the sequence such as

    x:=x....

and start it without writing the above. It is especially useful when You
want to perform an action on some long-named fields in Your object
identifier or the variables with quite big names. The main sequence is:

    <function>=

where <function> may be one of those: +, -, *, /, AND, &, OR, ||, <<, >>

example use:

DEF object:objectname
...
PROC increase(value)
    object.member+=value
    // same as: object.member:=object.member+value
ENDPROC

@endnode
@node CH_5A
@title "5A. format (;)"
5A. format (;)
--------------
As suggested in chapter 1A, a statement generally stands in its
own line, but several of them may be put together on one line
by separating them with semicolon, or one may be spread over
more than one line by ending each line in a comma ",". Examples:

a:=1; WriteF('hello!\\n')
DEF a,b,c,d,                      /* too many args for one line (faked) */
    e,f,g

statements may be:
- assignments
- conditional statements, for statements and the like, (see @{" 5E " link CH_5E }-5K)
- void expressions
- labels
- assembly instructions

The comma is the primary character to show that you do not wish to
end the statement with the next linefeed, but from v3 on, any token
that cannot legally be the end of a line causes the statement to
continue. Furthermore, if not all "[" and "(" occuring in a statement
have been closed off, a statement will continue also.
examples of such tokens:

+ - * / =
< >= <=
:= . <=> ::
{ [ (
AND OR BUT THEN IS              -> others too, but these are most useful

example of bracketing:

a:=[
    [1,2],
    [3,4]
   ]                    -> assignment ends here.

@endnode
@node CH_5B
@title "5B. statement labels and gotos (JUMP)"
5B. statement labels and gotos (JUMP)
-------------------------------------
Labels are global-scoped identifiers with a ':' added to them, as in:

mylabel:

they may be used by instructions such as JUMP, and to reference
static data. They may be used to jump out of all types of loops (although
this technique is not encouraged, (see @{" 5F " link CH_5F }, EXIT), but not out of
procedures. In normal E programs they are mostly used with inline assembly.
Labels are always globally visible.

Usage of JUMP:  JUMP <label>
continues execution at <label>. You are not encouraged to use this
instruction, it's there for situations that would otherwise increase
the complexity of the program. Example:

IF done THEN JUMP stopnow

/* other parts of program */

stopnow:

@endnode
@node CH_5C
@title "5C. assignment (:=)"
5C. assignment (:=)
-------------------
The basic format of an assignment is:   <var> := <exp>
Examples: a:=1,  a:=myfunc(),  a:=b*3

In E multiple variables can be assigned at once, when <exp> is
a function that returns multiple returnvalues. (see @{" 6G " link CH_6G })

@endnode
@node CH_5D
@title "5D. assembly mnemonics"
5D. assembly mnemonics
----------------------
In E, inline assembly is a true part of the language, they need not
be enclosed in special "ASM" blocks or the like, as is usual in
other languages, nor are separate assemblers necessary to assemble
the code. This also means that it obeys the E syntax rules, etc.
(see @{" 15A " link CH_15A } to read all about the inline assembler). Example:

DEF a,b
b:=2
MOVEQ  #1,D0             /* just use some assembly statements */
MOVE.L D0,a              /* a:=1+b  */
ADD.L  b,a
WriteF('a=\\d\\n',a)       /* a will be 3 */

@endnode
@node CH_5E
@title "5E. conditional statement (IF)"
5E. conditional statement (IF)
------------------------------
IF, THEN, ELSE, ELSEIF, ENDIF

syntax:         IF <exp> THEN <statement> [ ELSE <statement> ]
or:             IF <exp>
                  <statements>
                [ ELSEIF <exp>           /* multiple elseifs may occur */
                  <statements> ]
                [ ELSE ]
                  <statements>
                ENDIF

builds a conditional block. Note that there are two general forms of
this statement, a single-line and a multiple-line version.


@endnode
@node CH_5F
@title "5F. for-statement (FOR)"
5F. for-statement (FOR)
-----------------------
FOR, TO, STEP, DO, ENDFOR

syntax:         FOR <var> := <exp> TO <exp> STEP <step> DO <statement>
or:             FOR <var> := <exp> TO <exp> STEP <step>
                  <statements>
                ENDFOR

builds a for-block, note the two general forms. <step> may be
any positive or negative constant, excluding 0, and is optional. Example:

FOR a:=1 TO 10 DO WriteF('\\d\\n',a)

The body of FOR may contain CONT and EXIT statements with the following
syntax:

CONT <boolexp>

Will continue executing at the next iteration of the loop of the boolexp
is true.

EXIT <boolexp>

Which allows you to exit the loop if boolexp is true.


@endnode
@node CH_5G
@title "5G. while-statement (WHILE)"
5G. while-statement (WHILE)
---------------------------
WHILE, DO, ENDWHILE

syntax:         WHILE <exp> DO <statement>
or:             WHILE <exp>
                  <statements>
                ENDWHILE

builds a while-loop, which is repeated as long as <exp> is TRUE. Note
the one-line/one-statement version and the multiple line version.

WHILE may also contain CONT and EXIT statements, like FOR.

This great idea was brought by Martin Kuchinka - author of PowerD compiler0.
The WHILE-ENDWHILE loop is expanded with following keywords:

ELSEWHILE

Alternative loop part. Any number of ELSEWHIlE keywords can be used in each
WHILE loop. Usage is same as for WHILE:

ELSEWHILE <condition>

with only one exception from the rule. As for now it DOESN'T support "DO"
keyword (it means: You cannot write ELSEWHILE x DO y as it will cause an
error)

Every time the WHILE loop is executed, program searches for a condition
that returns true and executes proper part of program. In other case it
returns the loop

ALWAYS

This keyword is designed especially to be used together with ELSEWHILE.
It declares part of program that will be called every time the loop is
executed and at last one condition is true.

Sample code (as short as possible :):

PROC main()

    DEF a=10

    WHILE a=10
        a:=a+1
        WriteF('Incrementing "a" by 1\\n');
    ELSEWHILE a=11
        a:=0
        WriteF('Setting "a" to zero\\n');
    ALWAYS
        WriteF('Loop is called again now\\n');
    ENDWHILE

    WriteF('a=\\d\\n',a)
ENDPROC

will return:

Incrementing "a" by 1
Loop is called again now
Setting "a" to zero
Loop is called again now
a=0

NOTES
- The infinite loops are much easier to create :))
- Please don't use ALWAYS keyword without ELSEWHILE

@endnode
@node CH_5H
@title "5H. repeat-statement (REPEAT)"
5H. repeat-statement (REPEAT)
-----------------------------
REPEAT, UNTIL

syntax:         REPEAT
                UNTIL <exp>

builds a repeat-until block: it will continue to loop this block until
<exp>=TRUE. Example:

REPEAT
  WriteF('Do you really, really wish to exit this program?\\n')
  ReadStr(stdout,s)
UNTIL StrCmp(s,'yes please!')

REPEAT..UNTIL may also contain CONT and EXIT statements, like FOR be aware
though that executing a CONT on a REPEAT will not re-evaluate the UNTIL
expression so you could end up in an infinite loop.

@endnode
@node CH_5I
@title "5I. loop-statement (LOOP)"
5I. loop-statement (LOOP)
-------------------------
LOOP, ENDLOOP

syntax:         LOOP
                  <statements>
                ENDLOOP

builds an infinite loop (unless you use EXIT to break the loop)

@endnode
@node CH_5J
@title "5J. select-case-statement (SELECT)"
5J. select-case-statement (SELECT)
---------------------------------
SELECT, CASE, DEFAULT, ENDSELECT

syntax:         SELECT <exp>
                [ CASE <exp>
                  <statements> ]
                [ CASE <exp>
                  <statements> ]   /* any number of these blocks */
                [ DEFAULT
                  <statements> ]
                ENDSELECT

builds a select-case block. Various expressions will be matched against
the variable, and only the first matching block executed. If nothing
matches, a default block may be executed.

SELECT character
  CASE 10
    WriteF('Gee, I just found a linefeed\\n')
  CASE 9
    WriteF('Wow, this must be a tab!\\n')
  DEFAULT
    WriteF('Do you know this one: "\\c" ?\\n',character)
ENDSELECT

next to the old SELECT <exp> which works on expressions in the CASE
statements, E has a SELECT <maxrange> OF <exp> which works with constants
and or ranges of constants in a CASE. not only is this for many applications
more powerful, it is also much faster for large numbers of cases (>5
usually), or if cases are equally probable. It assumes however, that all
CASEs lie within a small integer range from 0 TO n-1, where n is something
reasonable, for example 10, or 256 for characters.

example:

SELECT 128 OF FgetC(stream)
  CASE "\\n","\\b"
    WriteF('line ending\\n')
  CASE "\\t"," "
    WriteF('whitepace\\n')
  CASE "0" TO "9"
    WriteF('Integer\\n')
  CASE "A" TO "Z", "a" TO "z", "_"
    WriteF('Identifier\\n')
  DEFAULT
    WriteF('some other character\\n')
ENDSELECT

DEFAULT will be hit not only for those for which there is no CASE,
but also for the chars that fall out of the range, i.e. 128 TO 255
(and >255, and <0).
note that speed costs: because this SELECT uses a jump-table
to quickly get at the right CASE, it'll use 2*<maxrange> bytes,
256 in this case.


@endnode
@node CH_5K
@title "5K. increase statement (INC/DEC)"
5K. increase statement (INC/DEC)
--------------------------------
INC, DEC

syntax:         INC <var>,n
                DEC <var>,n

short for <var>:=<var>+n and <var>:=<var>-n.

The second part of the instruction (,n) can be omitted and defaults to 1. In
this case they are equivalent to var++ and var-- and the only difference is
that these are statements, and do not return a value.


@endnode
@node CH_5L
@title "5L. void expressions (VOID)"
5L. void expressions (VOID)
---------------------------
VOID

syntax:         VOID <exp>

[obsolete: only for backwards compatability]
calculates the expression without the result going anywhere. Only useful
for a clearer syntax, as expressions may be used as statements without
VOID in E anyway. This may cause subtle bugs though, as "a:=1" assigns
a the value 1, but "a=1" as a statement will do nothing. E will give you
a warning if this happens.


@endnode
@node CH_5M
@title "5M. memory deallocation (END)"
5M. memory deallocation (END)
-----------------------------
END is the complement to NEW. any pointer obtained should be
deallocated (and only!) with END

        END a

or even

        END a,b,c

where the arguments are PTRs to some type. END frees the amount of space
that is being pointed to, so if a is PTR TO LONG it will only free 4
bytes. so if you allocated a with NEW a[NUM], free it with

        END a[NUM]

NIL-pointers may safely be passed to NEW. Pointers are also
nuked to NIL afterwards.
If a is a PTR to an object that is an instance of a class, END
will dynamically get the size to be freed from the class-object,
so if you have an object of class b which is 32 bytes, but the
pointer you're freeing with is a baseclass pointer (only 24
bytes), END will correctly deallocate 32 bytes. this only works
for classes.

You can imagine END p as a macro for:

        IF p
          p.end()
          FastDispose(p,p.classobject.virtuallen)
          p:=NIL
        ENDIF

If you don't call END on memory allocated by NEW, it will be deallocated
automatically at the end of the program, however no destructor methods
(.end()) will be called at this stage (i.e.  if this is necessary you'll
have to do it yourself).

note that the NEW and END make use of the FastNew() and FastDispose()
functions (described elsewhere) which work quite differently from
NewR() and Dispose(). (see @{" 9F " link CH_9F } for details).



@endnode
@node CH_5N
@title "5N. Generate an error or warning (FATAL/WARN)"
5N. Generate an error or warning (FATAL/WARN)
---------------------------------------------

You may wish to include warnings or error message into your code (usually
based on conditional compilation). The following statements will generate
an error or warning during compilation.

        FATAL 'this is a fatal error'
  
and

        WARN 'this is a warning'
  
Neither of these strings will end up in the compiled code and wil simply
be displayed to the user when compiling. A useful use case for this is to
display an error or warning when using the wrong E compiler.

  #ifndef EVO_3_4_0
    FATAL 'This code requires at least E-VO compiler v3.4.0'
  #endif

In addition to genrating warnings, the built in warnings can be
enabled and disabled using 

  WARN TRUE

  and

  WARN FALSE

Make sure you only place them around the parts of the code you want to
disable warnings for. If you accidentally forget to re-enable them, the
warnings will be disabled for the rest of the file.

@endnode
@node CH_5O
@title "5O. code and data sections (SECTION)"
5O. code and data sections (SECTION)
------------------------------------

USAGE
  SECTION type,mem

PARAMETERS
  Second parameter is optional. First one decides about the section type
  [CODE/DATA], second - memory type [none=PUBLIC, FAST or CHIP]

ABOUT
  This keyword allows a programmer creating programs splitted up to more
  than one hunk

NOTE
  - OPT LARGE should be set if You want to use it.
  - Although SECTION DATA allows storing code, please don't do it.
  - it is allowed only in executable files, not in modules

  and a small example - don't blame me for it, I want to make it as simple
  as possible ;)

  MODULE 'tools/pt', 'dos/dos'

  PROC main()
      pt_play({module})
      Wait(SIGBREAKF_CTRL_C)
      pt_stop()
  ENDPROC

      SECTION DATA,CHIP
  module:
      INCBIN 'pt_module'

@endnode

@node CH_5P
@title "5P. Negative conditions/loops (IFN/ELSEIFN etc)"
5P. Negative conditions/loops (IFN/ELSEIFN etc)
-----------------------------------------------

While it is perfectly possible to use the Not() operator in loops or
conditions to indicate the negative condition eg.

   REPEAT
   ..
   UNTIL Not(failed)

There are a set of commands which are shorter and generate more efficient
code. These commands are:

    UNTILN
    WHILEN
    ELSEWHILEN
    IFN
    ELSEIFN
    EXITN
    CONTN
    
They are used in exactly the same way as their counterparts eg.

    REPEAT
    ..
    UNTILN failed

    or

    IFN v
    ..
    ENDIF

Rather than inverting the results of the operation (which is what adding a
Not() would do and thus adding extra steps to the calculation these commands
simply alter the expected branch condition so its direct opposite and so are
more efficient.

@endnode

@node CH_6A
@title "6A. proc definition and arguments (PROC)"
6A. proc definition and arguments (PROC)
----------------------------------------

You may use PROC and ENDPROC to collect statements into your own functions.
Such a function may have any number of arguments, and several return values.

PROC, ENDPROC

syntax:         PROC <label> ( <args> , ... )
                ENDPROC <returnvalue>, ...

defines a procedure with any number of arguments. Arguments are of type LONG
or optionally of type PTR TO <type> (see @{" 8B " link CH_8B }) and need no further
declaration. The end of a procedure is designated by ENDPROC. If no
return value is given, 0 is returned. Example: write a function that
returns the sum of two arguments:

PROC add(x,y)         /* x and y are local variables */
ENDPROC x+y           /* return the result  */

a short version:

PROC add(x,y) IS x+y

You can also denote a procedure as SAFE

PROC SAFE add(x,y)..

This will ensure that registers D2-D7/A2-A6 are all saved on the stack
at the start of the procedure and restored at the end. Normally E will
automatically save any registers used by E generated code (register
variables or those allocated to registers by the optmimiser). If you
write your own ASM code however you may want tell E to store the
registers on the stack - of course you can do this in your own ASM code
but this is an easier way.

PROC GETA4 func()..

The GETA4 modifier for a PROC definition tells the E compiler to add a small
block of code to the start of the PROC call to restore the A4 register.
This is useful if the procedure is being called as a hook or as a new
thread.

PROC STACKCHECK func()..

Enables the STACKCHECK option for this procedure only. See the STACKCHECK
command line option for more details. This is useful for adding stack checking
to recursive function calls if you need to check to ensure that there is enough
stack space available. It will generate an "STCK" exception in the event that
the stack space is too low.

NOREGS - This is an extra option you can add on the end of a proc statement

eg.

PROC sub1() NOREGS

This will override the OPT REG in use and ensure that no registers are used
as variable storage during the duratioon of the procedure definition. This
may be useful if you have inline asm code that needs the registers to be
available.

There is also an extension of the ENDPROC command that allows you to
control what assembly instruction is used to return from the routine.

ENDPROC WITH RTR or ENDPROC WITH RTE

If you don't understand what this is for then you will never need this.
It is used to return from a system exception handler.

@endnode
@node CH_6B
@title "6B. local and global definitions: scope (DEF)"
6B. local and global definitions: scope (DEF)
---------------------------------------------
You may define additional local variables besides those which are
arguments with the DEF statement. The easiest way is simply like:

DEF a,b,c

declares the identifiers a, b and c as variables of your function.
Note that such declarations should be at the start of your function.

DEF

syntax:         DEF <declarations>,...

declares variables. A declaration has one of the forms:

          <var>
          <var>:<type>              where <type>=LONG,<objectident>,...
          <var>[<size>]:<type>      where <type>=ARRAY,STRING,LIST

(see @{" 8A " link CH_8A } for more examples, as that is where the types are introduced).
For now, we'll use the <var> form.
Arguments to functions are restricted to basic types; (see @{" 8B " link CH_8B }).
A declaration of a basic type can have an initialisation, in the current
version this must be an integer (not an expression):

DEF a=1,b=2

A program consists of a set of functions, called procedures, PROCs. Each
procedure may have Local variables, and the program as a whole may have
Global variables. At least one procedure should be the PROC main(), as
this is the module where execution begins. A simple program could look
like:

DEF a, b                            /* definition of global vars */

PROC main()                         /* all functions in random order */
  bla(1)
ENDPROC

PROC bla(x)
  DEF y,z                           /* possibly with own local vars */
ENDPROC

To summarize, local definitions are the ones you make at the start of
procedures, and which are only visible within that function. Global
definitions are made before the first PROC, at the start of your
source code, and they are globally visible. Global and local variables
(and of course local variables of two different functions) may have the
same name, local variables always have priority.

@endnode
@node CH_6C
@title "6C. endproc/return"
6C. endproc/return
------------------
As stated before, ENDPROC marks the end of a function definition, and may
return a value. Optionally RETURN may be used at any point in the function
to exit, if used in main(), it will exit the program.

RETURN [<returnvalue>]          /* optional */

Example:

PROC getresources()
  /* ... */
  IF error THEN RETURN FALSE  /* something went wrong, so exit and fail */
  /* ... */
ENDPROC TRUE   /* we got this far, so return TRUE */

a very short version of a function definition is:

PROC <label> ( <arg> , ... ) IS <exp>

("RETURN" instead of "IS" is allowed, but obsolete)
These are function definitions that only make small computations, like
faculty functions and the like:  (one-liners :-)

PROC fac(n) IS IF n=1 THEN 1 ELSE fac(n-1)*n


@endnode
@node CH_6D
@title "6D. the 'main' function"
6D. the 'main' function
-----------------------
The PROC called main is only of importance because it is called as first
function; it behaves exactly the same as other functions, and may also
have local variables. Main has no arguments: the command-line arguments
are supplied in the system-variable "arg", or can be checked with
ReadArgs() (dos.library function)

@endnode
@node CH_6E
@title "6E. built-in system variables"
6E. built-in system variables
-----------------------------
Following global variables are always available in your program,
they're called system variables.

arg             As discussed above, contains a pointer to a zero-terminated
                string, containing the command-line arguments. Don't use
                this variable if you wish to use ReadArgs() instead.
stdout          Contains a file-handle to the standard output (and input).
                If your program was started from the workbench, so no
                shell-output is available, WriteF() will open a
                CON: window for you and put its file handle here.
stdin           file-handle of standard input
conout          This is where that file handle is kept, and the console
                window will be automatically closed upon exit of your
                program. (see @{" 9A " link CH_9A }, WriteF(), on how to use these two
                variables properly).
execbase,       These four variables are always provided with their
dosbase,        correct values.
gfxbase,
intuitionbase
stdrast         Pointer to standard rastport in use with your program,
                or NIL. The built-in graphics functions like Line()
                make use of this variable.
wbmessage       Contains a ptr to a message you got if you started
                from wb, else NIL. May be used as a boolean to detect
                if you started from workbench, or even check any
                arguments that were shift-selected with your icon.
                See WbArgs.e in the Src/Args dir how to
                make good use of wbmessage.
exception       exception ID used in exception handling
exceptioninfo   additional information used in exception handling
thistask        address of this task
__pool          address of the memory pool (OS 3.0+ only and when pooling is
                in use).
utilitybase     base reference for utility.library (only initialised if
                OPT UTILIB is in use).

@endnode
@node CH_6F
@title "6F. default arguments to functions"
6F. default arguments to functions
----------------------------------

default arguments allows you to specify for one or more
arguments of a procedure which is the default value, if the
procedure is called with less args than parameters. for example,
a procedure like:

PROC bla(a,b=1,c=NIL)

can be called like:                     is equivalent with:

bla(a,b,c)                              bla(a,b,c)
bla(a,b)                                bla(a,b,NIL)
bla(a)                                  bla(a,1,NIL)

This can be useful and also express something about the
procedures function, i.e. that most of the time one would
call it with NIL anyway, so why not leave it out for clarity.
That's also why you should not overdo it with D.A.: do not
start specifying non-sensical values for procedures
out of pure laziness, if you feel a certain parameter really
has no default value.

to make calls with fewer args unambiguous, D.A. declarations
can only apply to the last 0..n parameters of a PROC of n parameters.

for example, illegal is:  PROC bla(a,b=1,c)

(you should then simply reorder the parameters, of course).
arguments supplied in a call are filled in from left to right,
missing arguments are added with D.A.'s as needed.

@endnode
@node CH_6G
@title "6G. multiple return values"
6G. multiple return values
--------------------------
In E you can return any number of return values (max 3 in
Amiga E because of implementation reasons). How?

RETURN <exp>,<exp>,<exp>                 (or ENDPROC, of course)

example:

PROC sincos(rad)
  DEF sin,cos
  /* whatever computation is needed */
ENDPROC sin,cos

call with:

s,c:=sincos(3.14)
s:=sincos(1.00)

as you can see, there's a new statement of the form:

<var> , ... := <exp>

where <exp> makes mostly only sense as function call.
note two things:
- you can decide yourself how many values you wish to receive.
  this makes sense when the first retval is the main one,
  and the second/third optional infos, which might only be
  important to some callers.
- this form is a _statement_. this means that when you would
  call sincos() as part of another expression, only the
  first (the regular) return value is used:   fun(sincos(1.0))


@endnode
@node CH_6H
@title "6H. function values"
6H. function values
-------------------
With v3, you can also have functions as values, and pass these
freely around. they're different from quoted expression, since
they're called just like normal PROCs. example:

fun:={myproc}                   -> get PROC ptr

...

fun(1,2,3)                      -> apply to args, just like normal PROC

notes:
- you have to be sure that the PROC you have a ptr to and the
  number of args are the same. the compiler can't check this for you.
- even worse: you have to be sure the ptr is a PROC at all.
  there is a compiler warning that may help you with this.



@endnode
@node CH_7A
@title "7A. const (CONST)"
7A. const (CONST)
-----------------
syntax:         CONST <declarations>,...

Enables you to declare a constant. A declaration looks like:
<ident>=<value>
constants must be uppercase, and will in the rest of the program be
treated as <value>. Example:

CONST MAX_LINES=100, ER_NOMEM=1, ER_NOFILE=2

CONST STR='test'
CONST F=3.5

You cannot declare constants in terms of others that are being
declared in the same CONST statement: put these in the next.

CONST, ENUM and SET declarations are always global, i.e. it is not
possible to declare constants local to a PROC. The best place for
constant declarations is at the top of your source, but E-VO also
allows you to put them between two PROCs, for example.

@endnode
@node CH_7B
@title "7B. enumerations (ENUM)"
7B. enumerations (ENUM)
-----------------------
Enumerations are a specific type of constant that need not be given values,
as they simply range from 0 .. n, the first being 0. At any given point
in an enumeration, you may use the '=<value>' notation to set or reset
the counter value. Example:

ENUM ZERO, ONE, TWO, THREE, MONDAY=1, TUESDAY, WEDNESDAY

ENUM ER_NOFILE=100, ER_NOMEM, ER_NOWINDOW

@endnode
@node CH_7C
@title "7C. sets (SET)"
7C. sets (SET)
--------------
Sets are again like enumerations, with the difference that instead of
increasing a value (0,1,2,...) they increase a bitnumber (0,1,2,...) and
thus have values like (1,2,4,8,...). This has the added advantage that
they may be used as sets of flags, as the keyword says.
Suppose a set like the one below to describe properties of a window:

SET SIZEGAD,CLOSEGAD,SCROLLBAR,DEPTH

to initialise a variable to properties DEPTH and SIZEGAD:

winflags:=DEPTH OR SIZEGAD

to set an additional SCROLLBAR flag:

winflags:=winflags OR SCROLLBAR

and to test if either of both of two properties hold:

IF winflags AND (SCROLLBAR OR DEPTH) THEN /* whatever */


@endnode
@node CH_7D
@title "7D. built-in constants"
7D. built-in constants
---------------------
Following are built-in constants that may be used:

TRUE,FALSE      Represent the boolean values (-1,0)
NIL             (=0), the uninitialised pointer.
ALL             Used with string functions like StrCopy() to copy all
                characters
GADGETSIZE      Minimum size in bytes to hold one gadget (see @{" 9D " link CH_9D },
                Gadget()) [obsolete]
OLDFILE,NEWFILE,READWRITE
                Mode-parameters for use with Open()
EMPTY           used with methods (might be keyword in the future)
STRLEN          Always has the value of the length of the last immediate
                string used. Example:

                Write(handle,'hi folks!',STRLEN)      /* =9 */
                
                The following built in constants are new and not active if
                you are using LEGACY mode.
TAG_DONE,TAG_END,TAG_IGNORE,
TAG_MORE,TAG_SKIP,TAG_USER
                Used with tag lists
OFFSET_BEGINNING,OFFSET_CURRENT,OFFSET_END
                Used with the dos.library Seek function


@endnode
@node CH_8A
@title "8A. about the 'type' system"
8A. about the 'type' system
---------------------------
E doesn't have a rigid type-system like Pascal or Modula2, it's even more
flexible than C's type system: you might as well call it a datatype-system.
This goes hand in hand with the philosophy that in E all datatypes are
equal: all basic small values like characters, integers etc. All have
the same 32bit size, and all other datatypes like arrays and strings
are represented by 32bit pointers to them. This way, the e compiler can
generate code in a very polymorphic way.
The (dis)advantages are obvious:

disadvantages of the E-type system
- less compiler checking on silly errors you make

advantages:
- low-level polymorphism, easier to make powerful generic functions.
- flexible way of programming: no problem that some types of return values
  don't match, no superfluous "casts" etc., no unnecessary errormessages.
- no hard to find errors when mixing data of different sizes in expressions


@endnode
@node CH_8B
@title "8B. the basic type (LONG/PTR)"
8B. the basic type (LONG/PTR)
-----------------------------
There's only one basic, non-complex variable type in E, which is the
32bit type LONG. As this is the default type, it may be declared as:

DEF a:LONG             or just:            DEF a

This variable type may hold what's known as CHAR/INT/PTR/LONG types in other
languages. A special variation of LONG is the PTR type. This type
is compatible with LONG, with the only difference that it specifies
to what type it is a pointer. By default, the type LONG is specified
as PTR TO CHAR. Syntax:

DEF <var>:PTR TO <type>

where type is either a simple type or a compound type. Example:

DEF x:PTR TO INT, myscreen:PTR TO screen

Note that 'screen' is the name of an object as defined in
intuition/screens.m

For example, if you open your own screen with:

myscreen:=OpenS(...   etc.

you may use the pointer myscreen as in 'myscreen.rastport'. However,
if you do not wish to do anything with the variable until you call
CloseS(myscreen), you may simply declare it as

DEF myscreen

Variable declarations may have optional initialisations, but only
integer constants, i.e. no full expression:

DEF a=1, b=NIL:PTR TO textfont


@endnode
@node CH_8C
@title "8C. the simple type (BYTE/CHAR/INT/WORD/LONG)"
8C. the simple type (BYTE/CHAR/INT/WORD/LONG)
---------------------------------------------
in the E language basic (single) variables cannot be 8 or 16 bit types
(BYTE/CHAR/INT/WORD); the reason for this must be clear by now.
However they may be used as data type to build ARRAYs from, set PTRs to,
use in the definition of OBJECTs etc. See those for examples.

The types are defined as follows:

     BYTE - Signed 8 bit values (-128 to 127)
     CHAR - Unsigned 8 bit values (0 to 255)
     INT  - Signed 16 bit values (-32768 to 32767)
     WORD - Unsigned 16 bit values (0-65535)
     LONG - Signed 32 bit values (-2147483648 to 2147483647)


@endnode
@node CH_8D
@title "8D. the array type (ARRAY)"
8D. the array type (ARRAY)
--------------------------
ARRAYs are declared by specifying their length (in elements):

DEF b[100]:ARRAY

this defines an array of 100 bytes. Internally, 'b' is a variable of
type LONG and a PTR to this memory area.
Default type of an array-element is CHAR, it may be anything by specifying:

DEF x[100]:ARRAY OF LONG
DEF mymenus[10]:ARRAY OF newmenu

where "newmenu" is an example of a structure, called OBJECTs in E.
Array access is very easy with:   <var>[<sexp>]

You can also just skip the "ARRAY OF" and use the following shortcut syntax:

DEF x[100]:LONG
DEF b[255]:CHAR

b[1]:="a"
z:=mymenus[a+1].mutualexclude

Note that the index of an array of size n ranges from 0 to n-1,
and not from 1 to n.
Note that ARRAY OF <type> is compatible with PTR TO <type>, with the
only difference that the variable that is an ARRAY is already
initialised.

Multi-dimensional arrays are also allowed and declared as follows:

DEF f[5][8][10]:ARRAY OF CHAR

The number of dimensions is not limited however the maximum size of the
array is 32767 elements in total (eg when all the dimensions are
multiplied together).

Accessing a multi-dimensional array is done using the same syntax:

f[1][3][5]:="a"



@endnode
@node CH_8E
@title "8E. the complex type (STRING/LIST)"
8E. the complex type (STRING/LIST)
----------------------------------
- STRINGs. Similar to arrays, but different in the respect that they may
  only be changed by using E string functions, and that they contain
  length and maxlength information, so string functions may alter them in a
  safe fashion, i.e: the string can never grow bigger than the memory
  area it is in. Definition:

  DEF s[80]:STRING

  The STRING datatype (called an estring) is backwards compatible with
  PTR TO CHAR and of course ARRAY OF CHAR, but not the other way around.
  (see @{" 9B " link CH_9B } on string functions for more details).

- LISTs. These may be interpreted as a mix between a STRING and an ARRAY
  OF LONG. I.e: this data structure holds a list of LONG variables which may
  be extended and shortened like STRINGs. Definition:

  DEF x[100]:LIST

  A powerful addition to this datatype is that it also has a 'constant'
  equivalent [], like STRINGs have ''. LIST is backward compatible with
  PTR TO LONG and of course ARRAY OF LONG, but not the other way around.
  (see @{" 9C " link CH_9C } and 2G) for more on this.


@endnode
@node CH_8F
@title "8F. the compound type (OBJECT)"
8F. the compound type (OBJECT)
------------------------------
OBJECTs are like a struct/class in C/C++ or a RECORD in pascal. Example:

OBJECT myobj
  a:LONG
  b:CHAR
  c:INT
ENDOBJECT

This defines a data structure consisting of three elements. Syntax:

OBJECT <objname>
  <membername> [ : <type> ]           /* any number of these */
ENDOBJECT

where type is one of the following:

CHAR/INT/LONG/<object>
PTR TO CHAR/INT/LONG/<object>
ARRAY OF CHAR/INT/LONG/<object>

(ARRAY with no type is short for ARRAY OF CHAR)

like DEF declarations, omitting the type means :LONG.

Note that <membername> need not be a unique identifier,
it may be in other objects too. There are lots of ways to use objects:

DEF x:myobj                      /* x is a structure */
DEF y:PTR TO myobj               /* y is just a pointer to it */
DEF z[10]:ARRAY OF myobj

y:=[-1,"a",100]:myobj            /* typed lists */

IF y.b="a" THEN /* ... */

z[4].c:=z[d+1].b++

(see @{" 4F " link CH_4F } and other parts of chapter 4 for these)

INT's or LONG ARRAYs in objects are by default put on even offsets and
ARRAY sizes are rounded to an even size:

OBJECT myobject
  len:CHAR, data[9]:ARRAY OF INT
  a:CHAR
ENDOBJECT

SIZEOF myobject is 22, and "data" starts at offset 2.

You can use the NOALIGN option on the object to disable this feature

OBJECT mystring NOALIGN
  len:CHAR, data[9]:ARRAY
ENDOBJECT

This will allow items to be placed on odd boundaries. This requires you have
chosen OPT 020 since the 68000 will crash if you try to access INT or LONG
items on odd boundaries.

'PTR TO' is the only type in OBJECTs that may refer to yet undeclared
other objects.

(see @{" 14A " link CH_14A } for all other OBJECT features that are somehow OO related)

@endnode
@node CH_8G
@title "8G. initialisation"
8G. initialisation
------------------

1. Always initialised to NIL (or else, if explicitly stated)
   - global variables
     NOTE: for documentation purposes, it's always nicer if you
     write =NIL in the definitions of variables that you expect to be NIL.
2. Initialised to '' and [] resp.
   - global and local STRINGs
   - global and local LISTs
3. Not initialised
   - local variables (unless explicitly stated)
   - global and local ARRAYs
   - global and local OBJECTs



@endnode
@node CH_8H
@title "8H. the essentials of the E typesystem"
8H. the essentials of the E typesystem
--------------------------------------
This section tries to explain how the E typesystem works from
another perspective.

Most problems people have while programming in E stem from their incorrect
view of how the E type-system works, Also, many people have an idea how
types work from their previous programming language, and try to apply this
to E, which is often fatal, because E is quite different when it come to
types.

The Type System.
but E is in essence a TYPELESS language. Indeed, variables may have a type,
but this is only used as a specification how to dereference a variable when
it is used as a pointer. In almost ALL other language constructions,
variables are treated as all being of the same type, namely the 32bit
typeless value.

In practise this means that for example in expressions with the exception
of the ".", "[]" and "++" operators etc., all operators and functions work
on 32bit values, regardless of whether they represent booleans, integers,
reals or pointers to something.

Pointer Types.
In the E type-system only 4 types exist, PTR TO CHAR, PTR TO INT,
PTR TO LONG and PTR TO <object>, where <object> is a name of a previously
defined OBJECT. When a variable (or an object member, as we'll see later)
is declared as being of this type, It means that if the variable contains
a value that is a legal pointer, this is how it should be dereferenced.

It is also possible to declare something as a PTR to a PTR. So for example

DEF c: PTR TO PTR TO CHAR

So c or c[n] would be a PTR but c[n][n] would be a CHAR

LONG, ARRAY etc.
All other types one may see in a DEF declaration are not really types, as
they really are only other ways of writing one of the above four. As an
example, ARRAY OF <type> is just another way of writing PTR TO <type>, with
the only difference that the former is automatically assigned the address
of an area of stackspace which is big enough to hold data for the #of
elements specified in square brackets.

Here's a table that shows all E 'types' in terms of the basic four:

ARRAY OF CHAR, ARRAY, STRING, LONG   (are equal to)   PTR TO CHAR
ARRAY OF INT                         (is equal to)    PTR TO INT
ARRAY OF LONG, LIST                  (are equal to)   PTR TO LONG
ARRAY OF <object>, <object>          (are equal to)   PTR TO <object>

- LONG is for variables that are not intended to be used as a pointer,
  i.e integers. Its equivalence with PTR TO CHAR is quite logical, as
  conceptually both talk about things that are measured in units of 1.
  (for example, "++" has the same effect on both)
- LIST and STRING are the same as their ARRAY equivalents, in respect
  to the fact that they're initialised to a piece of stack-space, but
  their stack representation is a little more complex to facilitate
  runtime bounds-checking (when used with the correct functions).
- an <object> is equivalent to [1]:ARRAY OF <object>. both represent
  an initialised PTR TO <object>.

In an OBJECT one can have the same declarations, with the addition of CHAR
and INT (similar to LONG), and the ommission of LIST and STRING, as these
are complex objects in their own right, and cannot be part of an object.

Deferencing.
Given a pointer p of some type,

"[]" may index other elements that are sequentially ordered next to
     the element it is currently pointing to. note that this allows for
     both positive and negative indices, and also no assumptions are made
     about where and how many elements are actually allocated.

"++" sets the pointer to the next element in memory, "--" to the previous
     one. note that these operators always operate on the pointer and
     never on the the element the pointer is pointing to.

"."  works similar to "[]", only now indexes the pointer by name, i.e. the
     pointer must be a PTR TO <object>.

"[]" and "." may be concatenated to a pointer p in any sequence, given the
fact that the previous resulting value again is known to be of a "PTR TO"
type.

One does not need to write out a de-reference in total, as in other
languages, e.g. if p is an ARRAY OF obj, instead of having to write
p[index].member you can write just p[index], which logically results
in the address of that object. This also explains why p[].member
is equivalent to p.member, since p[] is the same as p when it points
to an object.

Reference Semantics.
Another type-related issue that makes E somewhat different from other
languages and thus harder to grasp is it's accent on Reference Semantics
rather than Value Semantics. I'll try to argue why that's good here.

Informally, Reference Semantics means that objects in a language (mostly
other than the simple ones like LONGs) are represented by pointers, while
Value Semantics treats these objects as just being themselves. An example
of a language that has only Value Semantics is BASIC, examples of
languages that have them both are the C/C++ and Pascal type-of languages,
and examples of Reference only are newer Object Oriented languages,
functional languages like LISP and of course E.

Using Reference Semantics doesn't mean being occupied with pointers
all the time, rather you're worrying about them a lot less then in the
mixed case or the Value-only case, especially since in real life programs
most non-trivial data-structures get allocated dynamically which implies
pointers. The best example of this is LISP, where one programs heavily
with pointers without noticing. In E, one could easily forget STRING
is a pointer, given the easy by which one can pass it around to other
functions; in C often lots of "&" are needed where in the equivalent E
case none are, and the Oberon equivalent of bla('hallo') looks like
bla(sys.ADR('hallo')) because the string doesn't represent a pointer,
but a value as a whole...



@endnode
@node CH_9A
@title "9A. io functions"
9A. io functions
----------------

        WriteF(formatstring,args,...)
        PrintF(formatstring,args,...)

prints a string (which may contain formatting codes) to stdout. Zero
to unlimited arguments may be added. Note that, as formatstrings may
be created dynamically, no check on the correct number of arguments
is (can be) made. Examples:

WriteF('Hello, World!\\n')      /* just write a lf terminated string */

WriteF('a = \\d \\n',a)         /* writes: "a = 123", if a was 123 */

(see @{" 2F " link CH_2F } about strings for more).
NOTE: if stdout=NIL, for example if your program was started from the
Workbench, WriteF() will create an output window, and put the handle
in conout and stdout. This window will automatically be closed on
exit of the program, after the user typed a <return>. WriteF() is the
only function that will open this window, so if you want to do IO
on stdout, and want to be sure stdout<>NIL, perform a "WriteF('')"
as first instruction of your program to ensure output. If you want
to open a console window yourself, you may do so by placing the resulting
file handle in the 'stdout' and 'conout' variables, as your window will
then be closed automatically upon exit. If you wish to close this window
manually, make sure to set 'conout' back to NIL, to signal E that there's
no console window to be closed. PrintF() is the same as WriteF only
uses the v37+ buffered IO.
both return the length of the string that was printed.

        Out(filehandle,char)      and      char:=Inp(filehandle)

Either write or read one single byte to some file or stdout
if char=-1 then an EOF was reached, or an error occurred.
Out returns the number of bytes actually written (<>1 is error).
The functions use dos.library unbuffered IO, and as such are
slow for large amounts of data.

        len:=FileLength(namestring)

lets you determine the length of a file you *may* wish to load, and
also, if it exists (returns -1 upon error/file not found).

        ok:=ReadStr(filehandle,estring)

(see @{" 9B " link CH_9B })

        oldout:=SetStdOut(newstdout)
        oldin:=SetStdIn(newstdin)

Sets the standard output variable 'stdout' (not the value
returned from Input() and Output()!). Equivalent for:
oldout:=stdout; stdout:=newstdout. Same goes for the stdin
variable. 

        bool:=Eof(filehandle)
  
Checks if we are at the end of the file.

filehandle:=Fopen(name,mode)

Opens a file and returns the file handle. All open file handles are stored
in a global list that will automatically be closed when you program ends.

        Fclose(filehandle) 
 
Closes and open file and removes the handle from the internal list. You
should always use this rather than the OS Close() when using Fopen.
 
        PutF(filehandle, formatstr, args...)

Used in the same way as WriteF but writes the output to a filehandle rather
than to the standard output.
  
        blks:=ReadB(filehandle, blksize, numblocks, mem)

Read from an open file into memory. Reads a number of blocks of a given
block size and returns the number of blocks read.

        blks:=WriteB(filehandle, blksize, numblocks, mem)

Write to an open file from memory. Writes a number of blocks of a given
block size and returns the number of blocks written.

        len:=Size(filehandle)

Returns the length of a currently open file.

@endnode
@node CH_9B
@title "9B. strings and string functions"
9B. strings and string functions
--------------------------------

E has a datatype STRING. This is a string, from now on called 'Estring',
that may be modified and changed in size, as opposed to normal 'strings',
which will be used here for any zero-terminated sequence. Estrings are
downward compatible with strings, but not the other way around, so if an
argument requests a normal string, it can be either of them. If an Estring
is requested, don't use normal strings. Example of usage:

DEF s[80]:STRING, n                -> s is an estring with a maxlen of 80
ReadStr(stdout,s)                  -> read input from the console
n:=Val(s)                          -> get a number out of it
  -> etc.

Note that all string functions will handle cases where string tends to
get longer than the maximum length correctly;

DEF s[5]:STRING
StrAdd(s,'this string is longer than 5 characters',ALL)

s will contain just 'this '.
A string may also be allocated dynamically from system memory
with the function String(), (note: the pointer returned from this function
must always be checked against NIL)

        s:=String(maxlen)

DEF s[80]:STRING     is equivalent to     DEF s     and     s:=String(80)


        bool:=StrCmp(string,string,len=ALL)

compares two strings. len must be the number of bytes to compare,
or 'ALL' if the full length is to be compared. Returns TRUE or FALSE
(len is a default argument (see @{" 6F " link CH_6F }))

        StrCopy(estring,string,len=ALL)

copies the string into the estring. If len=ALL, all will be copied.
returns the estring.

        StrAdd(estring,string,len=ALL)

same as StrCopy(), only now the string is concatenated to the end.
returns the estring.

        StrAddChar(estring,char)

Same as StrAdd but adds a single character onto the end of an estring.

        len:=StrLen(string)

calculates the length of any zero-terminated string

        len:=EstrLen(estring)

returns the length of an estring

        max:=StrMax(estring)

returns the maximum length of a estring

  StringF(estring,fmtstring,args,...)

similar to WriteF, only now output goes to estring instead of stdout.
example:

StringF(s,'result: \\d\\n',123)

's' will be 'result: 123\\n'
returns the estring, and length as second returnvalue.

        AstringF(string,fmtstring,args,...)
  
array string version of StringF.. So no length checking is done so make sure
the output string is long enough or you will end up with memory corruption.


        RightStr(estring,estring,n)

fills estring with the last n characters of the second estring
returns the estring.

        MidStr(estring,string,pos,len=ALL)

copies any number of characters (including all if len=ALL) from
position pos in string to estring
NOTEZ BIEN: in all string related functions where a position in a
string is used, the first character in a string has position 0,
not 1, as is common in languages like BASIC.
returns the estring.

        value,read:=Val(string,read=NIL)

finds an integer encoded in ascii out of a string. Leading spaces/tabs
etc. will be skipped, and also hexadecimal numbers (1234567890ABCDEFabcdef)
and binary numbers (01) may be read this way if they are preceded by a
"$" or a "%" sign respectively. A minus "-" may indicate a negative integer.
Val() returns the number of characters read in the second argument, which
must be given by reference (<-!!!), or can be received as second
returnvalue. If "read" returns 0 (value will be 0 too) then the string did
not contain an integer, or the value was too sizy to fit in 32bit. "read"
may be NIL.

examples of strings that would be parsed correctly:
'-12345', '%10101010', '   -$ABcd12'

these would return both as "value" and in read a 0:
'', 'hello!'


        foundpos:=InStr(string1,string2,startpos=0)

searches string1 for the occurrence of string2, possibly starting from
another position than 0. Returned is the offset at which the substring
was found, else -1.

        foundpos:=InStri(string1,string2,startpos=0)

Same as InStr but with a case insensitive comparison.

        foundpos:=CharInStr(string,char,startpos=0)

Same as InStr but searches for a single character in string.

        foundpos:=CharInStri(string,char,startpos=0)

Same as InStri but searches for a single character in string.


        newstringadr:=TrimStr(string)

returns the *address* of the first character in a string, i.e., after
leading spaces, tabs etc.

        res:=EndsWith(string1,string2)

returns TRUE if a string1 ends with string2 otherwise returns FALSE
      
        UpperStr(string)     and      LowerStr(string)

changes the case of a string.
TAKE NOTE: these functions modify the contents of 'string', so they may
only be used on estrings, and strings that are part of your programs data.
Effectively this means that if you obtain the address of a string through
some amiga-system function, you must first StrCopy() it to a string of
your program, then use these functions.
they return the string.

        ch:=UpperChar(char)   and   ch:=LowerChar(char)
  
uppercase and lowercase functions that work on a single character.

        ok:=ReadStr(filehandle,estring)

will read a string (ending in ascii 10) from any file or stdout.
ok contains -1 if an error occurred, or an EOF was reached.
Note: the contents of the string read so far is still valid.
Also note that, like Inp() Out(), etc. this function makes use
of unbuffered 1.3 style IO, and thus may be slow. The dos.library
Fgets() function forms a nice alternative.

        SetStr(estring,newlen=ALL)

manually sets the length of a string. This is only handy when you read
data into the estring by a function other then an E string function,
and want to continue using it as an Estring.

The second parameter is optional and defaults to ALL which will
automatically calculate the length of a zero-terminated string.  For
example, after using a function that just puts a zero-terminated string
at the address of estring, use   SetStr(mystr) to make it manipulatable
again. If the string is too long, SetStr does nothing (this should always
be prevented).

        newstr:=StrClone(estring)
  
Creates a new estring that is a copy of the estring passed in. This is a
shorthand way of creating a new string and copying the contents of the
original estring.

        newstr:=AstrClone(string,len=ALL)
  
Similar to StrClone but creates a new estring that is a copy of the
string passed in. A partial string copy can be performed by passing in a
len value. This is a shorthand way of creating a new estring from an
existing string.
 
        AstrCopy(string1,string2,len=ALL)

'Array String Copy' copies string2 into the memory area denoted by string1.
string1 is typically not an estring but an ARRAY. size is the total #of
chars string1 can hold, i.e. if you write 5 and string2='helloworld',
string1 will be 'hell' + 0termination.

        AstpCopy(string1,string2,len=ALL)

This is based on the C stpcpy function. It is very similar to AstrCopy
except that AstpCopy() returns the address of the  end of the string dest
(that is, the address of the terminating zero byte) rather than the
beginning.

It is therefore optimised for concatenating several strings tgogether using
multiple calls to the function.

        order:=OstrCmp(string1,string2,max=ALL)

'Ordered String Compare' returns 1 if string1>string2, 0 for equal and -1
for less. only max chars are compared.

        StrIns(estring1,estring2,pos)

Inserts estring2 into estring1 at the specified pos. Returns the estring

        StrRem(estring,pos,len=ALL)
  
Removes a sub-string within an estring. len characters are removed starting
at position pos. The estring is returned.

        StrCompare(string1,string2,len=ALL)
        
Compares string1 and string2 for upto "len" characters and returns -1 if
string1<string2. 0 if string1 = string2 and +1 is string1 > string2 (note
that this is the opposite way around to OstrCmp)


for string linking functions (see @{" 9H " link CH_9H })


@endnode
@node CH_9C
@title "9C. lists and list functions"
9C. lists and list functions
----------------------------

Lists are like strings, only they consist of LONGs, not CHARs.
They may also be allocated either global, local or dynamic:

DEF mylist[100]:LIST         /* local or global */
DEF a
a:=List(10)                  /* dynamic */

(note that in the latter case, pointer 'a' may contain NIL)
Just as strings may be represented as constants in expressions, lists
have their constant equivalent:

[1,2,3,4]

The value of such an expression is a pointer to an already initialised list.
Special feature is that they may have dynamic parts, i.e, which will
be filled in at runtime:

a:=3
[1,2,a,4]

moreover, lists may have some other type than the default LONG, like:

[1,2,3]:INT
[65,66,67,0]:CHAR                    /* equivalent with   'ABC'   */
['topaz.font',8,0,0]:textattr
OpenScreenTagList(NIL,[SA_TITLE,'MyScreen',TAG_DONE])

As shown in the latter examples, lists are extremely useful with
system functions: they are downward compatible with an ARRAY OF LONG,
and object-typed ones can be used wherever a system function needs
a pointer to some structure, or an array of those.
Taglists and vararg functions may also be used this way.
NOTEZ BIEN: all list functions only work with LONG lists, typed-lists
are only convenient in building complex data structures and expressions.

As with strings, a certain hierarchy holds:
list variables -> constant lists -> array of long/ptr to long
When a function needs an array of long you might just as well give a list
as argument, but when a function needs a listvar, or a constant list,
then an array of long won't do.

It's important that one understands the power of lists and in particular
typed-lists: these can save you lots of trouble when building just
about any data-structure. Try to use these lists in your own programs,
and see what function they have in the example-programs.

summary:

[<item>,<item>,... ]          immediate list (of LONGs, use with listfuncs) 
[<item>,<item>,... ]:<type>   typed list (just to build data structures)

If <type> is a simple type like INT or CHAR, you'll just have the
initialised equivalent of ARRAY OF <type>, if <type> is an object-name,
you'll be building initialised objects, or ARRAY OF <object>, depending
on the length of the list.

If you write    [1,2,3]:INT   you'll create a data structure of 6 bytes,
of 3 16bit values to be precise. The value of this expression then
is a pointer to that memory area. Same works if, for example, you have
an object like:

OBJECT myobject
  a:LONG, b:CHAR, c:INT
ENDOBJECT

writing    [1,2,3]:myobject     would then mean creating a data structure
in memory of 8 bytes, with the first four bytes being a LONG with value 1,
the following byte a CHAR with value 2, then a pad byte, and the last
two bytes an INT (2 bytes) with value 3. you could also write:

[1,2,3,4,5,6,7,8,9]:myobject

you would be creating an ARRAY OF myobject with size 3. Note that such
lists don't have to be complete (3,6,9 and so on elements), you may
create partial objects with lists of any size

One last note on data size: on the amiga, you may rely on the fact that
a structure like 'myobject' has size 8, and that it has a pad byte
to have word (16bit) alignment. It is however very likely that an
E-compiler for 80x86 architectures will not use the pad byte and make
it a 7byte structure, and that an E-compiler for a sun-sparc architecture
(if I'm not mistaken) will try to align on 32bit boundaries, thus make
it a 10 or 12 byte structure. Some microprocessors (they are rare, but
they exist) even use (36:18:9) as numbers of bits for their types
(LONG:INT:CHAR), instead of (32:16:8) as we're used to. So don't make too
great an assumption on the structure of OBJECTs and LISTs if you want to
write code that stands a chance of being portable or doesn't rely on side
effects.

        ListCopy(listvar,list,num=ALL)

Copies num elements from list to listvar. example:
DEF mylist[10]:LIST
ListCopy(mylist,[1,2,3,4,5],ALL)
returns listvar.

        ListAdd(listvar,list,num=ALL)

Copies num items of list to the tail of listvar.
returns listvar.

        ListAddItem(listvar,value)

Adds a single item to the tail of listvar.
returns listvar.

        ListInsItem(listvar,value,pos)

Inserts a single item in a specific position of listvar.
returns listvar.

        ListRemItem(listvar,pos)

Remove a single item at a specific position of listvar.
returns listvar.

        ListSwapItem(listvar,pos1,pos2)

Swaps a two single itemd at specific positiond of listvar.
returns listvar.

        ListCmp(list,list,num=ALL)

Compares two lists, or some part of them.

        len:=ListLen(list)

Returns length of list, like     ListLen([a,b,c])    would return 3

        max:=ListMax(listvar)

returns maximum possible length of a listvar.

        value:=ListItem(list,index)

functions as    value:=list[index]    with the difference that
list may also be a constant value instead of a pointer. This is
very useful in situations like this where we directly want to
use a list of values:

WriteF(ListItem(['ok!','no mem!','no file!'],error))

this prints an errormessage according to "error". it's similar to:

DEF dummy:PTR TO LONG
dummy:=['ok!','no mem!','no file!']
WriteF(dummy[error])

        list:=ListClone(oldlist)
  
creates a new list with the same specifications as the old list and copies
across the values in the old list. The values are just blindly treated as
LONG's so if they are references to objects then the new list will reference
the same objects. 

        SetList(listvar,newlen)

manually sets the length of a list. This will only be useful when you read
data into the list by a function other then a list-specific function,
and want to continue using it as a true list.

for list functions that make use of quoted expressions (see @{" 11C " link CH_11C }).
for list linking functions (see @{" 9H " link CH_9H }).


@endnode
@node CH_9D
@title "9D. intuition support functions"
9D. intuition support functions
-------------------------------

        wptr:=OpenW(x,y,width,height,IDCMP,wflags,title,
                    screen,sflags,gadgets,taglist=NIL)

creates a window where wflags are flags for window layout
(like BACKDROP, SIMPLEREFRESH e.d, usually $F) and sflags are
for specifying the type of screen to open on (1=wb,15=custom).
screen must only be valid if sflags=15, else NIL will do.
gadgets may point to a glist structure, which you can easily
create with the Gadget() function, else NIL.

        CloseW(wptr)

closes that screen again. Only difference from CloseWindow()
is that it accepts NIL-pointers and sets stdrast back to NIL.

        sptr:=OpenS(width,height,depth,sflags,title,taglist=NIL)

opens a custom screen for you. depth is number of bitplanes (1-6, 1-8 AGA).

        CloseS(sptr)

as CloseW(), now for screens.

        nextbuffer:=Gadget(buffer,glist,id,flags,x,y,width,string)

[obsolete]
This function creates a list of gadgets, which can then be put in your
window by giving them as an argument to OpenW(), or afterwards with
intuition functions like AddGlist().
buffer is mostly an ARRAY of at least GADGETSIZE bytes to hold all the
structures associated with one gadget, id is any number that may help you
remember which gadget was pressed when an IntuiMessage arrives.
Flags are: 0=normal gadget, 1=boolean gadget, 3=boolean gadget that is
selected. Width is width in pixels, that should be large enough to hold
the string, which will be auto-centered. glist should be NIL for the first
gadget, and glistvar for all others, so E may link all gadgets.
The function returns a pointer to the next buffer (=buffer+GADGETSIZE).
Example for three gadgets:

CONST MAXGADGETS=GADGETSIZE*3

DEF buf[MAXGADGETS]:ARRAY, next, wptr

next:=Gadget(buf,NIL,1,0,10,20,80,'bla')   /* the 1st gadget */
next:=Gadget(next,buf,... )
next:=Gadget(next,buf,... )                /* any amount linked 2 1st */

wptr:=OpenW( ...,buf)

        code:=Mouse()

gives you the current state of all 2 or 3 mouse buttons; left=1,
right=2 and middle=4. If for example code=3 then left and right were
pressed together.
NOTEZ BIEN: this is not a real intuition function, if you want to
know about mouse-events the proper way, you'll have to check the
intuimessages that your window receives. This is the only E
function that directly checks the hardware, and thus only useful
in demo-like programs and for testing. (DO NOT USE THIS FUNCTION
IN PROGRAMS THAT ARE SUPPOSED TO WORK UNDER THE OS)

        bool:=LeftMouse(win)     WaitLeftMouse(win)

intuition Mouse() alternatives for programs that only want to test for
a mouseclick.

        x:=MouseX(win)     y:=MouseY(win)

[obsolete]
enables you to read the mouse coordinates. win is the window
they need to be relative to.

        class:=WaitIMessage(window)

This function makes it easier to just wait for a window event. It simply
waits until a intuimessage arrives, and returns the class of the event.
It stores other variables like code and qualifiers as private global
variables, for access with functions described below.
WaitIMessage() represents the following code:

PROC waitimessage(win:PTR TO window)
  DEF port,mes:PTR TO intuimessage,class,code,qual,iaddr
  port:=win.userport
  IF (mes:=GetMsg(port))=NIL
    REPEAT
      WaitPort(port)
    UNTIL (mes:=GetMsg(port))<>NIL
  ENDIF
  class:=mes.class
  code:=mes.code             /* stored internally */
  qual:=mes.qualifier
  iaddr:=mes.iaddress
  ReplyMsg(mes)
ENDPROC class

as you see, it gets exactly one message, and does not forget about
multiple messages arriving in one event, if called more than once.
For example, say you opened a window that displays something and just
waits for a closegadget (you specified IDCMP_CLOSEWINDOW only):

WaitIMessage(mywindow)

or, you have a program that waits for more types of events, handles
them in a loop, and ends on a closewindow event:

WHILE (class:=WaitIMessage(win))<>IDCMP_CLOSEWINDOW
  /* handle other classes */
ENDWHILE

        code:=MsgCode()    qual:=MsgQualifier()    iaddr:=MsgIaddr()

These all supply you with the private global variables as mentioned
before. the values returned are all defined by the most recent call
to WaitIMessage(). Example:

IF class:=IDCMP_GADGETUP
  mygadget:=MsgIaddr()
  IF mygadget.userdata=1 THEN  /* ... user pressed gadget #1 */
ENDIF



@endnode
@node CH_9E
@title "9E. graphics support functions"
9E. graphics support functions
------------------------------

All graphics support functions that do not explicitly ask for a rastport,
make use of the system-variable 'stdrast'. It is automatically defined by
the last call to OpenW() or OpenS(), and is set to NIL by CloseW() and
CloseS(). Calling these routines while stdrast is still NIL is legal.
stdrast may be manually set by SetStdRast() or stdrast:=myrast

        Plot(x,y,colour=1)

Draws a single dot on your screen/window in one of the colours available.
colour ranges from 0-255, or 0-31 on pre-AGA machines.

        Line(x1,y1,x2,y2,colour=1)

Draws a line

        Box(x1,y1,x2,y2,colour=1)

Draws a box

        Colour(foreground,background=0)

sets the colours for all graphics functions (from the library) that
do not take a colour as argument. This is the colour *register*
(i.e 0-31) and not colour *value*
NOTE: functions that have "colour" as an argument, change the Apen
of stdrast.

        TextF(x,y,formatstring,args,...)

exactly the same function as WriteF(), only outputs to some (x,y) on
your stdrast, instead of stdout. (see @{" 9A " link CH_9A }, WriteF() and strings in the
language reference). returns the length of the resulting string.

        oldrast:=SetStdRast(newrast)

changes the output rastport of the E graphics functions

        SetTopaz(size=8)

[obsolete: don't use]
lets you set the font of the rastport "stdrast" to topaz. size is 8 or 9.
Only to be used as last resort if you can't support font-sensitivity.

        SetColour(screen,colourreg,r,g,b)

set colour register (0..255) of screen to certain RGB values.
each rgb value has a range of 0..255, i.e. 24bit colour. this
function will automatically rescale to 12bit colour if no AGA
is present, and also use the correct function.


@endnode
@node CH_9F
@title "9F. system support functions"
9F. system support functions
----------------------------

        bool:=KickVersion(vers)

Will give TRUE if the kickstart in the machine your program is running
on is equal or higher than vers, else FALSE

        mem:=New(n)

This dynamically creates an array (or memory area, if you wish) of
n bytes. Difference with AllocMem() is that it is called automatically
with flags $10000 (i.e cleared mem, any type) and that no calls to
Dispose() are necessary, as it is linked to a memory list that is
automatically de-allocated upon exit of your program.
(see @{" 4K " link CH_4K }, also)

        mem:=NewR(n)

same as New(), only now automatically raises the "MEM" exception
instead of return if no memory could be allocated.

        mem:=NewM(n,flags)

same as NewR(), but also allows you to specify flags (MEMF_CHIP etc.)

        Dispose(mem)

Frees any mem allocated by New(). You only have to use this function
if you explicitly wish to free memory _during_ your program, as all
is freed at the end anyway.

        CleanUp(returnvalue=0)

[obsolete: use exception handling]
Exits the program from any point. It is the replacement for the DOS
call Exit(): never use it! instead use CleanUp(), which allows
for the deallocation of memory, closing libraries correctly etc.
The return value will be given to dos as returncode.
Cleanup() is ONLY necessary if you have to exit at a point different
from ENDPROC in main.

        amount:=FreeStack()

returns the amount of free stack space left. This should always be 1000 or
more. (see @{" 16C " link CH_16C } on how E organizes its stack. If you don't do heavy
recursion, you need not worry about your free stack space.

Stack space can also be monitored further using these calls

        amount:=UsedStack()
        amount:=TotalStack()

which will give details of the total stack space available and used by the
current task. The total stack will not match the value specified in the code
exactly as some stack space is reserved automatically on top of the size
requested.

        bool:=CtrlC()

Returns TRUE if Ctrl-C was pressed since you last checked, else FALSE.
This only works for programs running on a console, i.e. cli-programs.

Also CtrlD(), CtrlE(), CtrlF() work the same way but for each of the
possible break conditions.

Example of how these last three functions may be used:

/* calculate faculty from command-line argument */

OPT STACK=100000

PROC main()
  DEF num,r
  num:=Val(arg,{r})
  IF r=0 THEN WriteF('bad args.\\n') ELSE WriteF('result: \\d\\n',fac(num))
ENDPROC

PROC fac(n)
  DEF r
  IF FreeStack()<1000 OR CtrlC() THEN CleanUp(5)    /* xtra check */
  IF n=1 THEN r:=1 ELSE r:=fac(n-1)*n
ENDPROC r

Of course, this recursion will hardly run out of stack space, and when it
does, it's halted by FreeStack() so fast you won't have time to press
CtrlC, but it's the idea that counts here.
A definition of fac(n) like:

PROC fac(n) IS IF n=1 THEN 1 ELSE fac(n-1)*n

would be less safe.


     mem:=FastNew(size)      FastDispose(mem,size)   FastDisposeList(list)

FastNew() and FastDispose() are replacements for NewR(size) and
Dispose(ptr) (they are used by NEW and END). this is what they have
in common:
- "MEM" exceptions may be Raised
- memory is always cleared
- auto-dealloc at end of program
- NIL accepted in deallocation
but the following differences should be noted (positive):
- they are varying from 10 to 50 times faster (!)
- they use way less memory for small objects
- they do not fragment memory
[all this is for objects <=256 bytes, for bigger ones NewR() and
 Dispose() are used].
negative:
- they do not free mem, but recycle it.
- FastDispose() needs exact size of allocation. END also.

List allocated with NEW need the function FastDisposeList(). Because
Lists have a length, the size parameter isn't needed.

        GetA4()
  
The register A4 is used in E for accessing global variables. If A4 gets
corrupted by any code (particularly inline assembler) then it will not be
possible to access any global variables. GetA4() will restore the correct
value in A4. However the value is stored in a way that is not re-entrant
so it is not suitable for code that needs to be made resident (eg library
code). Using this function will cause a compiler error if you have OPT PURE
turned on.

        mem:=Alloc(size)
  
This is used in conjuntion with OPT POOL to allocate memory from the pool.
Memory pools are available only in OS v39 and above. 0 will be returned in
when not enough free pool space is available. The size of the pool is set
in the OPT POOL command.

        Free(mem)
  
Releases memory allocated with the Alloc command back to the pool.

        MemFill(addr,size,value)
        
Fills the memory at addr to addr+size-1 with value. The value can be 0-255.

@endnode
@node CH_9G
@title "9G. math and other functions"
9G. math and other functions
----------------------------

        a:=And(b,c)           a:=Or(b,c)           a:=Not(b)
        a:=Eor(b,c)

These work with the usual operations, boolean as well as arithmetic.
Note that for And() and Or() an operator exists.

        a:=Mul(b,c)           a:=Div(a,b)

Performs the same operation as the '*' and '/' operators, but now in
full 32bit. For speed reasons, normal operations are 16bit*16bit=32bit
and 32bit/16bit=16bit. This is sufficient for nearly all calculations,
and where it's not, you may use Mul() and Div(). NOTE: in the Div
case, a is divided by b, not b by a.

        bool:=Odd(x)          bool:=Even(x)

Return TRUE or FALSE if some expression is Odd or Even

        Min(a,b)        Max(a,b)

compute min and max of the two given ints.

        randnum:=Rnd(max)       seed:=RndQ(seed)

Rnd() computes a random number from an internal seed in range 0 .. max-1.
For example,  Rnd(1000)   returns an integer from 0..999
To initialise the internal seed, call Rnd() with a negative value;
the Abs() of that value will be the initial seed.

RndQ() computes a random number "Q"uicker than Rnd() does, but returns
only full range 32bit random numbers. Use the result as the seed for
the next call, and for startseed, use any large value, like $A6F87EC1

        absvalue:=Abs(value)

computes the absolute value.

        s:=Sign(v)

computes the sign of v, i.e. returns -1,0,1

        a,b:=Mod(c,d)

Divides 32bit c by 16bit d and returns 16bit modulo a and optionally
a 16bit result of the division b. If these 16bit limits are exceeded,
Mod() will not give correct results.

        x:=Shl(y,num)         x:=Shr(y,num)

arithmatically shifts y num bits to left or right (set new bits to 0).

        x:=Lsl(y,num)         x:=Lsr(y,num)

logically shifts y num bits to left or right (set new bits to 0).

        x:=Rol(y,num)         x:=Ror(y,num)

rotates y num bits to left or right.


        a:=Long(adr)          a:=Int(adr)          a:=Word(adr)
        
        a:=Char(adr)          a:=Byte(adr)

peeks into memory at some address, and returns the value found. This
works with 32, 16 and 8 bit values and signed or unsigned according to the
respective types. Note that the compiler does not check if 'adr' is valid.
These functions are available in E for those cases where reading and writing
in memory with PTR TO <type> would only make a program more complex or less
efficient. You are _not_ encouraged to use these functions.

NOTE: In E-VO versions prior to 3.6.0 (and other E languages based directly
from the original E source) Int() did not sign extend the results. This
has been corrected in v3.6.0 of E-VO but you can restore the old functioning
by specifying OPT LEGACYINT

        PutLong(adr,a)       PutInt(adr,a)        PutChar(adr,a)

Pokes value 'a' into memory. See: Long() in addition:

        PutByte(adr,a) and PutWord(adr,a)
        
are also defined but are just aliases for PutChar and PutInt since there is
no actual difference in functionality when the value is signed or not, the
value is simply saved to that address.

        y:=Bounds(x,a,b)

makes sure x lies between bounds a and b, and adjust acordingly if
necessary. 

It equals:    y:=IF x<a THEN a ELSE IF x>b THEN b ELSE x

        b:=Chk(x)
  
Returns FALSE if x is 0 and TRUE for any other value.

  r:=Compare(a,b)
  
Does a comparison between a and b and returns -1 if a < b, 0 if a=b and +1
if a>b

  r:=Ucompare(a,b)
  
As above but treating a and b as unsigned 32 bit values

  r:=Fcompare(a,b)

As above but treating a and b as floating point values
  
  r:=StrCompare(a,b,len)

As above but performing a string comparison on a and b.
(see @{" 9B " link CH_9B } on string functions for more details)

  r:=MemCompare(a,b,len)

This is a direct memory compare function. It is essentially the same as the
StrCompare function except that it will not stop if it its the NIL termiator
character found at the end of a string.



@endnode
@node CH_9H
@title "9H. string and list linking functions"
9H. string and list linking functions
-------------------------------------
E provides for a set of functions that allows the creation of
linked list with the STRING and LIST datatype, or strings and lists
that were created with String() and List() respectively. As you may
know by now, strings and lists, complex datatypes, are pointers
to their respective data, and have extra fields to a negative offset
of that pointer specifying their current length and maxlength. the
offsets of these fields are PRIVATE. as an addition to those two,
any complex datatype has a 'next' field, which is set to NIL by
default, which may be used to build linked list of strings, for example.
in the following, I will use 'complex' to denote a ptr to a STRING
or LIST, and 'tail' to denote another such pointer, or one that already
has other strings linked to it. 'tail' may also be a NIL pointer,
denoting the end of a linked list.
[note: these String-list functions have nothing to do with E-lists or
 Lisp-Cell lists]
The following functions may be used:

        complex:=Link(complex,tail)

puts the value tail into the 'next' field of complex. returns again complex.
example:

DEF s[10]:STRING, t[10]:STRING
Link(s,t)

creates a linked list like:    s --> t --> NIL

        tail:=Next(complex)

reads the 'next' field of var complex. this may of course be NIL, or
a complete linked list. Calling Next(NIL) will result in NIL, so it's
safe to call Next when you're not sure if you're at the end of a linked
list.

        tail:=Forward(complex,num)

same as Next(), only goes forward num links, instead of one, thus:

Next(c) = Forward(c,1)

You may safely call Forward() with a num that is way too large;
Forward will stop if it encounters NIL while searching links, and
will return NIL.

        DisposeLink(complex)

same as Dispose(), with two differences: it's only for strings and
lists allocated with String() or List(), and will automatically
de-allocate the tail of complex too. Note that large linked lists
containing strings allocated with String() as well as some allocated
locally or globally with STRING may also be de-allocated this way.

For a good example of how linked lists of strings may be put to
good use in real-life programs, see Src/Utils/D.e


@endnode
@node CH_9I
@title "9I. lisp-cells and cell functions"
9I. lisp-cells and cell functions
---------------------------------
yep. that's right. you thought LISP was fun, then try E now.
[or: the story about why E is a better LISP than LISP itself :-)]

Starting from v3, E has the cell datatype, almost identical to cells
in the LISP language. more technically, E has:

'Conservative Mark and Sweep Garbage-Collected Lisp-Cells'

basically this amounts to being able to allocate LISP-cells,
which are pairs of two values:

        x:=<a|b>

which is much like NEW [a,b]:LONG, only now E will automatically
deallocate the 8 bytes in question itself when it finds out it needs
memory and no pointers are pointing to the cell. In practise this
means that you can freely have functions create cells as temporaries,
without worrying about freeing them. And any LISP-programmer will
be able to explain to you that with cells you can build any
data-structure (most notably trees and lists). [note: this text
does not thorougly explain how to make full use of cells, since
dozens of LISP books have been written about this]

Selecting the values can easily be done using Car(x) and Cdr(x),
two LISP-functions which select head and tail (first and second)
element of the cell. if x is a PTR TO LONG, even x[0] and x[1]
are allowed.

One can also write lists of cells:

        <a,b,c>

(note the commas) as short for

        <a|<b|<c|NIL>>>

An alternative for selection with Car/Cdr is E's unification:

        x <=> <a,b|c>
        a+b+c

instead of:

        Car(x)+Car(Cdr(x))+Cdr(Cdr(x))

lisp-cell unification resembles E-list unification (see @{" 4L " link CH_4L }). for
example:

        x <=> <1,2|a>

equals:

        IF Car(x)=1
          IF Car(Cdr(x))=2
            a:=Cdr(Cdr(x))
            ...


A lisp-nil value is available "<>", which equals "NIL" and "0", but not
"[]".

some functions are available (note that Cons() is _only_ available
through <...>)

        h:=Car(c)       t:=Cdr(c)

get the head and the tail value of a cell c

        bool:=Cell(c)

gives a bool for whether or not c points at a cell, so Cell(<1>)=TRUE,
and Cell(3.14)=FALSE. This is not a fast function.

        n:=FreeCells()

tell you about the amount of free cells available. very slow function.
there should be no need to call this function other than curiosity.

        SetChunkSize(k)

Sets the chunksize to allocate for cells to k kilobyte. default is 128k.
This function can only be called once, and only before the first cons
(<..>) allocation takes place. Thereafter it has no effect.

In general, get a good book about lisp to understand more about
programming with cells.

One can write any LISP-functions in E, with exactly the same functionality:

PROC append(x,y) IS IF x THEN <Car(x)|append(Cdr(x),y)> ELSE y
PROC nrev(x) IS IF x THEN append(nrev(Cdr(x)),<Car(x)>) ELSE NIL
PROC sum(x) IS IF x THEN Car(x)+sum(Cdr(x)) ELSE 0

using a destructive implementation for functions like these is
also allowed.

techy stuff:
E's garbage collector implements a conservative mark and sweep algorithm
that was tested to be 5 to 25 times faster than several logical and
functional language implementations on the Amiga. Conservative
means that in case of doubt, the GC will not deallocate a cell. this
is necessary since in a typeless language such as E, the GC can
easily bump into a value that is not a valid pointer etc.

The GC allocates big chunks (default 128k), in which it allocates
cells. if out of cells, it will collect garbage by scanning the
stack and registers for pointers into the cellspace, and recursively
mark them. after that, all unmarked cells are reused, and if the
gain after a collection was only small, a new chunk is allocated
(if this fails, "MEM" is raised).

interaction with other E values:
- storing other values in cells is no problem whatsoever. objects,
  strings, floats, anything can be put into a cell without confusing
  the GC too much.
- storing cells in other values, for example a ptr to a cell in a
  dynamic object, is problematic, since the GC won't be able to find
  it there. a solution for this will be provided. However I think this
  case will seldom occur.
  ptrs to cells can safely be stored in global and local variables,
  even registers, and any stack datastructures.
  [and most importantly, in other cells!]

caveats:
- The GC currently can't collect cells that have a Car-list >1000
  long or so, i.e. <<<NIL:a>:b>:c>, but then 1000 instead of 3
  entries. this will hardly ever occur since lists like this
  are usually formed as Cdr-lists, which the GC can handle into
  infinity. (it will raise "STCK" if this fails")
- inline assembly code should never push stuff on the stack that
  is not LONG aligned. this was already necessary in v2.1b,
  but now is even more essential.

There's a trade off in chunk-size between time and space.
Allocating small chunks obviously is nice since you won't waste
any memory, however, when collecting garbage, the effort for
each pointer to trace is almost proportional to the number of
spaces. therefore:
- if speed is most important tune the chunkspacesize such that
  that only one space is needed. if the top cell-memory usage at a
  certain time is 50k, a chunkspace of 100k or 150k will give
  optimal performance.
- if memory usage is more important, in the example above a
  chunksize of 20k or 30k will be quite optimal for memory.
In general, time a heavy usage of your cell-algorithm with
different sizes to see what trade-off suits you best.



@endnode
@node CH_9J
@title "9J. Boopsi functions"
9J. Boopsi functions
--------------------

E-VO Provides a number of functions that are useful when working with Boopsi
objects. These are normally found in amiga.lib but since E-VO doesn't
provide any way to link with amiga.lib these have been implemented as
function calls.

These are particularly useful when working with the new Reaction objects
that are introduced in OS3.2

All of these boopsi functions are only available when targeting OS 37 or
above.

        res:=DoMethod(object, methodid, ...)
        res:=DoMethodA(object, message)

These invoke the specific message on the given object. The only difference
is the way the parameters are passed. DoMethod expects a stream of
parameters and DoMethodA expects the message to be a taglist starting with
the methodid code and followed by any additional parameters and ending with
TAG_DONE

        res:=CoerceMethod(class, object, methodid, ...)
        res:=CoerceMethodA(class, object, message)

Function invokes the supplied message on the specific object as though it
were the specified class

        res:=DoSuperMethod(class, object, methodid, ...)
        res:=DoSuperMethodA(class, object, message)

Function invokes the supplied message on the specified object though as it
were the superclass of the specified class

        Set(object, attr, value, ...)
        Sets(object, attr, value)

Assign a value assigned to specified attribute of the object

        res:=Get(object, attr, store)
        res:=Gets(object, attr)

Ask specified object for a value assigned to specified attribute

@endnode
@node CH_10A
@title "10A. built-in library calls"
10A. built-in library calls
---------------------------
As you may have noticed from previous sections, the piece of code
automatically linked to the start of your code, called the "initialisation
code", always opens the three libraries intuition, dos, graphics and (and
sometimes mathieeesingbas), and because of this, the compiler has all the
calls to those four libraries (including exec) integrated in the compiler
(there are a few hundred of them). These are up to AmigaDos v3.1 (v40). To
call Open() from the dos library, simply say:

handle:=Open('myfile',OLDFILE)

or AddDisplayInfo() from the graphics library:

AddDisplayInfo(mydispinfo)

it's as simple as that.


@endnode
@node CH_10B
@title "10B. interfacing to the amiga system with the v47 modules"
10B. interfacing to the amiga system with the v47 modules
----------------------------------------------------------
To use any other library than the five in the previous section, you'll
need to resort to modules. Also, if you wish to use some OBJECT or CONST
definition from the Amiga includes as is usual in C or assembler,
you'll need modules. Modules are binary files which may include constant,
object, library and function (code) definitions. The fact that they're
binary has the advantage over ascii (as in C and assembly), that they
need not be compiled over and over again, each time your program is
compiled. The disadvantage is that they cannot simply be viewed; they
need a utility like ShowModule (see @{" 17A " link CH_17A }) to make their contents
visible. The modules that contain the library definitions (i.e the calls)
are in the root of emodules: (the modules dir in the distribution), the
constant/object definitions are in the subdirectories, structured just
like the originals from Commodore.

MODULE

syntax:         MODULE <modulenames>,...

Loads a module. A module is a binary file containing information on
 libraries, constants, and sometimes functions. Using modules enables you to
use libraries and functions previously unknown to the compiler.

Now for an example, below is a short version of the
source/examples/asldemo.e source that uses modules to put up a filerequester
from the 2.0 Asl.library.


MODULE 'Asl', 'libraries/Asl'

PROC main()
 DEF req:PTR TO filerequester
 IF aslbase:=OpenLibrary('asl.library',37)
  IF req:=AllocFileRequest()
   IF RequestFile(req) 
    WriteF('File: "\\s" in "\\s"\\n',req.file,req.drawer)
   ENDIF
   FreeFileRequest(req)
  ENDIF
  CloseLibrary(aslbase)
 ENDIF
ENDPROC


From the modules 'asl', the compiler takes asl-function definitions like
RequestFile(), and the global variable 'aslbase', which only needs to be
initialised by the programmer (i.e it should not be defined by the
programmer).  From 'libraries/Asl', it takes the definition of the
filerequester object, which we use to read the file the user picked.  Well,
that wasn't all that hard:  did you think it was that easy to program a
filerequester in E?

Note that when objects imported from modules contain references to objects
from yet other modules, with an object member of type PTR TO <object>, this
type will only be set correctly if the object in question is already
available, otherwise it is set to LONG.  The only way to influence this is
rearranging the order of modules in the MODULE statement.


@endnode
@node CH_10C
@title "10C. compiling own modules"
10C. compiling own modules
--------------------------
with v3, you can gather all PROCs, CONSTs, OBJECTs and to
some extent also global variables that you feel somehow belong
together in one source, write "OPT MODULE" to signal E-VO that
this is supposed to be a module, and then compile all to a
.m file to be used in the main program, just like you used to
do with the old modules.

by default, all elements in a module are PRIVATE, i.e. not accessable
to the code that imports the .m file. to show which elememts
you wish to be visible in the module, simply write EXPORT
before it:

EXPORT ENUM TESTING,ONE,TWO,THREE,FOUR

EXPORT DEF important_glob_var, bla:PTR TO x

EXPORT OBJECT x
  next, index, term
ENDOBJECT

EXPORT PROC burp()
  /* whatever */
ENDPROC

EXPORT lab:

"EXPORT" is useful in making a distinction between private and
public, especially when all functions of an OBJECT can be accessed
via PROCs, you may wish to keep OBJECT private as an effective
method of data hiding. more on this topic, (see @{" 14C " link CH_14C })
[EXPORT can be written on any line, doesn't affect everything though.]

If in a module _all_ elememts need to be exported (for example
one with only constants), a 'OPT EXPORT' will export all, without
the need for individual EXPORT keywords.

global variables require extra attention:
- try to avoid global variables. having lots of globals
  in modules makes projects messy and error-prone
- globals in a module which are not exported function as
  local for the module, i.e. they'll never clash with globals
  in other modules. those who _are_ exported though, are
  combined with the others, i.e. if in both the main program
  and in a module a variable with the same name are used,
  this will be one and the same variable. that's why
  one can write DEF a[10]:ARRAY OF x in the main program,
  and EXPORT DEF a:PTR TO x in the module, to share the
  array. Also, if both use for example 'gadtools.m',
  only one of the two needs to initialise 'gadtoolsbase'
  for both to be able to make calls to the library.
  If you do not want librarybases to be shared (i.e. you
  want to have a local, private library base), simply
  redeclare it in a DEF that is not EXPORTed in the module.
  If you export a variable in a general purpose
  module, make sure to give it a pretty unique name.
  variables are combined in the main program, if you want to
  reference an exported variable from another module,
  you will have to export it there too.
- using globals in modules which provide general purpose
  datatypes needs special attention, as the module may be
  in use from more than one other module, in which case
  it may be unclear who is responsable for resources.
  take good care of this.


Module startup and module cleanup code

It is possible to declare one or more PROC in a module that will
automatically be executed during the program initialisation and cleanup.

PROC x() MODSETUP
...
ENDPROC

PROC x() MODCLEANUP
..
ENDPROC

The MODSETUP and MODCLEANUP tells the compiler that these procedures are to be
called automatically. This allows you to initialise any variables during
the program startup and free up resources at the end.


Using modules in modules

This requires little extra attention. If the module (B) you include
in your own module (A) is one that only declares CONSTs, LIBRARYs and
OBJECTs (without code) nothing special happens, however if B includes
PROCs, then it's obvious this code needs to be linked later to the
main program when A is linked. Therefore if a main program uses A,
B will need to be present for compilation. The fact that A needs
B is stored in A, and can be viewed with ShowModule. This chain
of uses may grow out to a tree of dependencies, which has the result
that if you use just one module in your program, a lot of others
are automatically linked to it. Thus, E's module system automatically
keeps track of dependancies that other languages need makefiles for.
E-VO also allows for circular inclusions, and loads/links modules at most
once (i.e. doesn't link unused modules). One thing E's module system
doesn't automatically do for you is recompile dependant modules. If
you change B, it is often necessary to recompile A too, since it
might be referring to offsets etc. in the old version of B, which
might cause code to crash. If this gets too complex in your project,
use a utility such as E-Build (see @{" 17I " link CH_17I }).


Try out the new ShowModule (see @{" 17A " link CH_17A }) to see what E-VO puts in modules.


Including modules from other directories.

By default, a module name is prefixed by 'emodules:' to obtain
the actual file. Now you can prefix the name with a '*' to
denote the directory the source is in, so:

MODULE 'bla', '*bla'

if this statement would be in source 'WORK:E/burp.e', these two
modules would be 'emodules:bla.m' and 'WORK:E/bla.m'.

This is naturally the way to include components of your app into
other parts. If you write modules that you use in many of your programs
it would be handy to store them in the emodules hierarchy, and the
place for this is the 'emodules:other/' dir.


Circular dependancies.

When writing sets of own modules, one should try to avoid
circular dependancies (E-VO does not enforce this), as in some
cases this can result in chicken and egg problems (you can't
recompile everything from scratch, you always need atleast
one readily compiled module).


@endnode
@node CH_10D
@title "10D. the modulecache"
10D. the modulecache
--------------------
(see @{" 17D " link CH_17D }, ShowCache/FlushCache about this).



@endnode
@node CH_11A
@title "11A. quoting and scope"
11A. quoting and scope
----------------------
Quoted expressions start with a backquote. The value of a quoted
expression is not the result from the computation of the expression,
but the address of the code. This result may then be passed on as
a normal variable, or as an argument to certain functions.
example:

myfunc:=`x*x*x

myfunc is now a pointer to a 'function' that computes x^3 when evaluated.
These pointers to functions are very different from normal PROCs, and
you should never mix the two up. The biggest differences are that a
quoted expression is just a simple expression, and thus cannot have its
own local variables. In our example, "x" is just a local or global variable.
That's where we have to be cautious:
if we evaluate myfunc somewhat later in the same PROC, x may be local,
but if myfunc is given as parameter to another PROC, and then evaluated,
x needs of course to be global. There's no scope checking on this.

@endnode
@node CH_11B
@title "11B. Eval()"
11B. Eval()
-----------
        Eval(func)

simply evaluates a quoted expression (exp = Eval(`exp)).

NOTE: because E is a somewhat typeless language, accidentally writing
"Eval(x*x)"  instead of  "Eval(`x*x)"  will go unnoticed by the
compiler, and will give you big runtime problems: the value of x*x
will be used as a pointer to code.

To understand why 'quoted expressions' is a powerful feature think of the
following cases: if you were to perform a set of actions on a set of
different variables, you'd normally write a function, and call that function
with different arguments. But what happens when the element that you want to
give as argument is a piece of code? in traditional languages this would not
be possible, so you would have to 'copy' the blocks of code representing
your function, and put the expression in it. Not in E. say you wanted to
write a program that times the execution time of different expressions. In
E you would simply write:

PROC timing(func,title)
  /* do all sorts of things to initialise time */
  Eval(func)
  /* and the rest */
  WriteF('time measured for \\s was \\d\\n',title,t)
ENDPROC

and then call it with:

timing(`x*x*x,'multiplication')
timing(`sizycalc(),'large calculation')


in any other imperative language, you would have to write out
copies of timing() for every call to it, or you would have to
put each expression in a separate function. This is just a simple
example: think about what you could do with data structures (LISTs)
filled with unevaluated code:

drawfuncs:=[`Plot(x,y,c),`Line(x,y,x+10,y+10,c),`Box(x,y,x+20,y+20,c)]

Note that this idea of functions as normal variables/values is not new
in E, quoted expressions are literally from LISP, which also has the
somewhat more powerful so-called Lambda function, which can also be
given as argument to functions; E's quoted expressions can also be
seen as parameterless (or global parameter only) lambdas.


@endnode
@node CH_11C
@title "11C. built-in functions"
11C. built-in functions
-----------------------

        MapList(varadr,list,listvar,func)

performs some function on all elements of list and returns all
results in listvar. func must be a quoted expression (see @{" 11A " link CH_11A }),
and var (which ranges over the list) must be given by reference. Example:

MapList({x},[1,2,3,4,5],r,`x*x)       results r in:     [1,4,9,16,25]

returns listvar.

        ForAll(varadr,list,func)

Returns TRUE if for all elements in the list the function (quoted
expression) evaluates to TRUE, else FALSE. May also be used to perform
a certain function for all elements of a list:

ForAll({x},['one','two','three'],`WriteF('example: \\s\\n',x))

        Exists(varadr,list,func)

As ForAll(), only this one returns TRUE if for any element the function
evaluates to TRUE (<>0). note that ForAll() always evaluates all elements,
but Exists() possibly does not.

        SelectList(v,list,listvar,quotedexp)

Much like MapList(), only now doesn't store the result from quotedexp,
it uses it as a boolean value, and only those values for which it is
true are stored in listvar (which should be capable of holding
the same amount of elements as list. example:

SelectList({x},[1,2,0,3,NIL],r,`x<>0)

results in r being [1,2,3].
returns length of listvar.

        SelectFirst(v,list,quotedexp)
        
This works exactly the same way as SelectList above but rather than
taking a listvar and returning all the items that match it simply
returns the first one that matches. If none are matched then it will
return NIL.


Example of how to use these functions in a practical fashion:
we allocate different sizes of memory in one statement, check them
all together at once, and free them all, but still only those that
succeeded. (example is v37+)


PROC main()
  DEF mem[4]:LIST,x
  MapList({x},[200,80,10,2500],mem,`AllocVec(x,0))           -> alloc some
  WriteF(IF ForAll({x},mem,`x) THEN 'Yes!\\n' ELSE 'No!\\n')     -> suxxes ?
  ForAll({x},mem,`IF x THEN FreeVec(x) ELSE NIL)  -> free only those <>NIL
ENDPROC


Note the absence of iteration in this code. Just try to rewrite this
example in any other language to see why this is special.



@endnode
@node CH_12A
@title "12A. float values"
12A. float values
-----------------
REALs (or FLOATs, whatever) are very different in E than in other
languages. This mainly has to do with the fact that E doesn't
really discriminate between types of values. One is advised to
understand these chapters _well_ before attempting to use floats.

In E, a float is just another 32bit value. The E compiler treats
them just like integers or pointers, with the difference that
their bit representation means something different. The E float
format is the IEEEsingle standard.

A float value looks like an integer value with the exception that
somewhere a "." is present. for example, the following are valid 
floats:

        3.14159    .1    1.    -12345.6

these aren't:

        .    1234

(i.e. atleast one "." and one "0-9" char must be present).

You can use these values at almost all places where LONG values
are legal, i.e. if you have have a function or datastructure that
handles arbitrary LONG values, it will also handle floats.

        DEF f=3.14
        myobj.x:=.1
        fun(f,2.73)



@endnode
@node CH_12B
@title "12B. computing with floats"
12B. computing with floats
--------------------------
Because to E a float will seem like just another LONG, it will happily
apply integer math to it when used in an expression, which is mostly
not what you want. Also, one would like to be able to convert to
integer and vice-versa. The float operator "!" handles all this.

assume in the following examples that a,b,c contain integer values,
and x,y,z float values.

By default, an expression in E is considered an integer expression.
what the "!" does when it occurs in an expression is the following:
- changes the expression from int to float. any operators following
  (+ * - / = <> > < >= <=) will be float operations. "!" may occur
  any number of times in an expression, changing from and to float
  again and again.
- the expression that did occur before the "!", if any, is converted
  to the appropriate type.

examples:

        x:=a!

converts "a" to float, and stores the result in x. "a" is an integer exp,
which is then toggled to float, which implies a conversion.

        a:=!x!

converts "x" to integer and stores the result in a.

        x:=y

        x:=Ftan(y)

no "!" is needed here since no operator-math or conversions are necessary.

        x:=!y*z

the "*" acts on y and z as floats, since "!" denotes the whole as
a float-exp. the float result is stored in x

        a:=b!*x+y!

a more complex example: the int "b" is converted to float, then x
and y are float-multiplied and float-added to it. The result is converted
to int and stored in the int "a"

        x:=!y*z-z*y+(a!)+z/z

        z:=!x*Fsin(!x*y)

all (+ * - /) are computed as float, and the int "a" is converted
to float somewhere in the middle. since "(" ")" denotes a new expression,
it has it's own status of "!". Same idea for the function below.

        IF !x<0.1 THEN WriteF('Float value too small!\\n')

as you can see, "!" also works on the six comparison operators.



@endnode
@node CH_12C
@title "12C. builtin float functions"
12C. builtin float functions
----------------------------

Some trans-math functions are present, more will probably follow.

        x:=Fsin(y)    x:=Fcos(y)    x:=Ftan(y)

usual sin() etc. functions. they work with radians.
New in v3.3 are Fatan() Fasin() Facos() Fsincos()
Fsinh() Fcosh() and Ftanh() which work analogously to
the functions above. Ftieee() and Ffieee() convert FFP
floats to and from ieee floats (as used by E).

        x:=Fabs(y)

compute absolute value of y

        x:=Ffloor(y)    x:=Fceil(y)

compute lowest and highest whole-number float value near y

        x:=Fexp(y)    x:=Flog(y)    Flog10(y)    x:=Fpow(y,z)    Fsqrt(y)

compute e^y, ln(y), log base 10, z^y and square root of y, respectively.

        x,n:=RealVal(s)

parses string "s" to produce float value x. will skip leading spaces and
tabs. n is the number of characters parsed from the start of the string,
or 0 if it couldn't be parsed as a float value. "x" will then be 0.0.
accepts negative numbers (and number without a ".", even).
example:

RealVal(' 3.14 ')    results in    3.14, 5
RealVal('blabla')    results in    0.0, 0

        s:=RealF(s,x,n)

Format a float value x to the estring s, with n positions after the ".".
max for "n" is 8, even less if you have lots of digits leading the ".".
an "n" of 0 will denote no fraction. The string is returned as result,
so it can be reused in a WriteF() for example:

WriteF('float = \\s\\n',RealF(s,3.14159,4)) results in  'float = 3.1416\\n'

RealF() tries hard to make sensible roundings for a certain "n", as
the example shows. negative numbers are also handled properly.

RealF(s,-3.14159,0)    results in    '-3'



@endnode
@node CH_12D
@title "12D. float implementation issues"
12D. float implementation issues
--------------------------------
As said before, the E float format is IEEE (single), this means that
older float code using the FFP format with the SpXxx functions will
have to be rewritten (as stated in the v2.1b docs). The mathffp
library is no longer directly supported by E starting from v3.0, and
you'll have to open this library like all others if you want to use it.

single IEEE's were chosen because:
- double IEEE's don't fit in a LONG
- the FFP format routines do not make use of a 68881 if present,
  the IEEE ones do. Furthermore the FFP format is incompatible
  with the 68881, which also uses IEEE format.
- IEEE is the worldwide float-format standard, which encourages
  data-file compatability among software/platforms.

E's float routines use the mathieeesingbas.library and the
mathieeesingtrans.library, which are not by default supplied
with the ancient v1.3 of the OS. This means that if you want
to write under / support 1.3 AND you want to use the _builtin_
floats, you have to make sure these libraries are present (they
seem to be available, maybe through commodore?).

Both E-VO and the programs it generates do not open these libraries
as long as no float-features are used.

If all else fails, one can always use other floatlibraries to use
floats with 1.3. I might recommend the tools/longreal.m module
which uses doubles.

In the future, E-VO will probably allow mathieee library calls to
be replaced with inline 68881 code transparently.



The bug in mathieeesingbas.library.

v3.1 (v40) of the amiga operating system is known to contain a bug in the
IEEE code of the divide and multiply functions.  If you're using E floats
under this version of the OS you will need to run a patch that fixes this
otherwise you may provoke crashes.  Even worse, even if you program under
another version of the OS, if your program uses these float operations, you
will have to notify your users to use this patch, otherwise your programs
may fail on other peoples machines.  

Another suitable patch is for example util/boot/PatchMathSB10.lha on Aminet
(or CD #7).  As far as E-VO itself is concerned, if E-VO has to use one of
these operations (for example for float constants) it will warn you that you
are not running the patch.



@endnode
@node CH_13A
@title "13A. defining exception handlers (TRY/CATCH or HANDLE/EXCEPT)"
13A. defining exception handlers (TRY/CATCH or HANDLE/EXCEPT)
-------------------------------------------------------------

E now has two styles of exception handling which you can happily mix and
match according to your needs.

The first (and original) exception mechanism in E is basically the
same as in ADA; it provides for flexible reaction on errors in your
program and complex resource management.

NOTE: the term 'exception' in E has very little to do with exceptions
caused directly by 680x0 processors.

An exception handler is a piece of program code that will be invoked
when runtime errors occurs, such as windows that fail to open or
memory that is not available. You, or the runtime system itself,
may signal that something is wrong (this is called "raising an
exception"), and then the runtime-system will try and find the
appropriate exception handler. I say "appropriate" because a program
can have more than one exception handler, on all levels of a program.
A normal function definition may (as we all know) look like this:

PROC bla()
  /* ... */
ENDPROC

a function with an exception handler looks like this:

PROC bla() HANDLE
  /* ... */
EXCEPT
  /* ... */
ENDPROC

The block between PROC and EXCEPT is executed as normal, and if no
exception occurs, the block between EXCEPT and ENDPROC is skipped, and
the procedure is left at ENDPROC. If an exception is raised, either
in the PROC part, or in any function that is called in this block,
an exception handler is invoked.

The second style of exception handling is more flexible and allows for
more than one exception handler in a single PROC. It is more akin to
the exception handling in more modern languages. The syntax for this is

  TRY
    ..
    do something
    ..
  CATCH
    ..
    handle the exception
    ..
  ENDTRY

So you can define a block of code anywhere within a procedure that can
handle exceptions raised within the scope of that block. Exception blocks
can be nested and in effect work exactly the same but this gives the
developer a greater level of control.

@endnode
@node CH_13B
@title "13B. using the Raise() function"
13B. using the Raise() function
-------------------------------
There are many ways to actually "raise" an exception, the simplest
is through the function Raise():

        Raise(exceptionID=0)

the exception ID is simply a constant that defines the type of
exception, and is used by handlers to determine what went wrong.
Example:

ENUM NOMEM,NOFILE  /* and others */

PROC bla() HANDLE
  DEF mem
  IF (mem:=New(10))=NIL THEN Raise(NOMEM)
  myfunc()
EXCEPT
  SELECT exception
    CASE NOMEM
      WriteF('No memory!\\n')
    /* ... and others */
  ENDSELECT
ENDPROC

PROC myfunc()
  DEF mem
  IF (mem:=New(10))=NIL THEN Raise(NOMEM)
ENDPROC

The "exception" variable in the handler always contains the value of
the argument to the Raise() call that invoked it.
In both New() cases, the Raise() function invokes the handler of
function bla(), and then exits it correctly to the caller of bla().
If myfunc() had its own exception-handler, that one would be invoked
for the New() call in myfunc(). The scope of a handler is from the start
of the PROC in which it is defined until the EXCEPT keyword, including
all calls made from there.

This has three consequences:
A. handlers are organized in a recursive fashion, and which handler is
   actually invoked is dependant on which function calls which at runtime;
B. if an exception is raised within a handler, the handler of a lower
   level is invoked. This characteristic of handlers may be used
   to implement complex recursive resource allocation schemes with
   great ease, as we'll see shortly.
C. If an exception is raised on a level where no lower-level handler
   is available (or in a program that hasn't got any handlers at all),
   the program is terminated. (i.e: Raise(x) has the same effect as
   CleanUp(0))

other functions:

        Throw(exceptionID,value)

same as Raise(), only now it takes an arbitrary value with it. in
a handler one can then scrutinize this value with the variable
'exceptioninfo'

        ReThrow()

has no args. simply does a Throw() on the current exception value,
IFF it is <>0.


@endnode
@node CH_13C
@title "13C. defining exceptions for built-in functions (RAISE/IF)"
13C. defining exceptions for built-in functions (RAISE/IF)
---------------------------------------------------------
With exceptions like before, we have made a major gain over the
old way of defining our own "error()" function, but still it is
a lot of typing to have to check for NIL with every call to New().

The E exception handling system allows for definition of exceptions
for all E functions (like New(), OpenW() etc.), and for all Library
functions (OpenLibrary(), AllocMem() etc.), even for those
included by modules. Syntax:

RAISE <exceptionId> IF <func> <comp> <value> , ...

the part after RAISE may be repeated with a ",".
Example:

RAISE NOMEM IF New()=NIL,
      NOLIBRARY IF OpenLibrary()=NIL

the first line says something like: "whenever a call to New() results
in NIL, automatically raise the NOMEM exception".
<comp> may be any of = <> > < >= <=
After this definition, we may write all through our programs:

mem:=New(size)

without having to write:

IF mem=NIL THEN Raise(NOMEM)

Note that the only difference is that 'mem' never gets any value
if the runtime system invokes the handler: code is generated for
every call to New() to check directly after New() returns and call
Raise() when necessary.

We'll now be implementing a small example that would be complex to solve
without exception handling: we call a function recursively, and in each
we allocate a resource (in this case memory), which we allocate before,
and release after the recursive call. What happens when somewhere high
in the recursion a severe error occurs, and we have to leave the program?
right: we would (in a conventional language) be unable to free all the
resources lower in the recursion while leaving the program, because all
pointers to those memory areas are stored in unreachable local variables.
In E, we can simply raise an exception, and from the end of the handler
again raise an exception, thus recursively calling all handlers and
releasing all resources. Example:


CONST SIZE=100000
ENUM NOMEM  /* ,... */

RAISE NOMEM IF AllocMem()=NIL

PROC main()
  alloc()
ENDPROC

PROC alloc() HANDLE
  DEF mem
  mem:=AllocMem(SIZE,0)         /* see how many blocks we can get */
  alloc()                       /* do recursion */
EXCEPT DO
  IF mem THEN FreeMem(mem,SIZE)
  ReThrow()                     /* recursively call all handlers */
ENDPROC


This is of course a simulation of a natural programming problem that
is usually far more complex, and thus the need for exception handling
becomes far more obvious. For a real-life example program whose error
handling would have become very difficult without exception
handlers, see the 'D.e' utility source.

The "DO" after an EXCEPT means that instead of jumping to ENDPROC,
the main code will simply continue execution in the handler as soon
as it gets there. it also sets exception to 0.
This is handy if you free resources local to a PROC in the handler.



@endnode
@node CH_13D
@title "13D. use of exception-ID's"
13D. use of exception-ID's
--------------------------
In real life an exception-ID is of course a normal 32-bit value,
and you may pass just about anything to an exception handler: for
example, some use it to pass error-description strings

Raise('Could not open "gadtools.library"!')

However, if you want to use exceptions in expandabele fashion and you
want to be able to use future modules that raise exceptions not defined
by your program, follow the following guidelines:

- Use and define ID 0 as "no error" (i.e. normal termination)

- For exceptions specific to your program, use the ID's 1-10000.
  Define these in the usual fashion with ENUM:

  ENUM OK,NOMEM,NOFILE,...

  (OK will be 0, and others will be 1+)

- ID's 12336 to 2054847098 (these are all identifiers
  consisting of upper/lowercase letters and digits of length 2,3 or 4
  enclosed in "") are reserved as common exceptions. A common exception
  is an exception that need not be defined in your program, and that
  may be used by implementors of modules (with functions in them) to
  raise exceptions: for example, if you design a set of procedures that
  perform a certain task, you may want to raise exceptions. As you would
  want to use those functions in various programs, it would be
  unpractical to have to coordinate the IDs with the main program,
  furthermore, if you use more than one set of functions (in a module,
  in the future) and every module would have a different ID for
  'no memory!', things could get out of hand.
  This is where common exceptions come in: the common out-of-memory
  ID is "MEM" (including the quotes): any implementor can now simply

  Raise("MEM")

  from all different procedures, and the programmer that uses the module
  only needs to suply an exception handler that understands "MEM"

  future modules that contain sets of functions will specify what
  exception a certain procedure may raise, and if these overlap
  with the IDs of other procedures, the task of the programmer
  that has to deal with the exceptions will be greatly simplified.

  examples:

  (system)

  "MEM"         out of memory
  "FLOW"        (nearly) stack overflow
  "STCK"        garbage collector has stack problems or stack checking failure
  "INDX"        array index out of range
  "^C"          Control-C break
  "ARGS"        bad args

  (exec/libraries)

  "SIG"         could not allocate signal
  "PORT"        could not create messageport
  "LIB"         library not available
  "ASL"         no asl.library
  "UTIL"        no utility.library
  "LOC"         no locale.library
  "REQ"         no req.library
  "RT"          no reqtools.library
  "GT"          no gadtools.library (similar for others)

  (intuition/gadtools/asl/gfx)

  "WIN"         failed to open window
  "SCR"         failed to open screen
  "REQ"         could not open requester
  "FREQ"        could not open filerequester
  "GAD"         could not create gadget
  "MENU"        could not create menu(s)
  "FONT"        problem getting font

  (dos)

  "OPEN"        could not open a file / file does not exist
  "OUT"         problems while writing
  "IN"          problems while reading
  "EOF"         unexpected end of file
  "FORM"        input format error
  "SEG"         loadseg problems

  The general tendency is:
  * all uppercase for general system exceptions,
  * mixed case for exceptions used by >1 app, but not general enough.
  * all lowercase for exceptions raised within your own
    multi-module application

- all others (including all negative IDs) remain reserved.



@endnode
@node CH_14A
@title "14A. OO features in E"
14A. OO features in E
---------------------
The features descibed here in this chapter are grouped as such
since they constitute what is generally seen as the three essential
main components that make a language 'Object Oriented' (i.e.
inheritance - data hiding - polymorhism). However in E they are
by no means a 'separate chapter' since each can be used in any
way with other E features.


@endnode
@node CH_14B
@title "14B. object inheritance"
14B. object inheritance
-----------------------
it's always annoying not being able to express dependencies between
OBJECTs, or reuse code that works on a particular OBJECT with a bigger
OBJECT that encapsulates the first. Object Inheritance allows you
to do just that in E. when you have an object a:

OBJECT a
  next, index, term
ENDOBJECT

you can make a new object b that has the same properties
as a (and is compatible with code for a):

OBJECT b OF a
  bla, x, burp
ENDOBJECT

is equivalent to:

OBJECT b
  next, index, term         /* from a */
  bla, x, burp
ENDOBJECT

with DEF p:b, you can directly not only access p.bla as usual,
but also p.next.

as an example, if one would have a module with an OBJECT to
implement a certain datatype (for example a doubly-linked-list),
and PROCs to support it, one could simply inherit from it, adding
own data to the object, and use the _existing_ functions to
manipulate the list. However, it's only in combination with
methods (descibed below), inheritance can show its real power.


@endnode
@node CH_14C
@title "14C. data hiding (EXPORT/PRIVATE/PUBLIC)"
14C. data hiding (EXPORT/PRIVATE/PUBLIC)
----------------------------------------
E has a very handy data-hiding mechanism. Other languages, like C++,
use data-hiding on classes, which raises the need for kludges (like
'friends'), and makes datahiding insecure (Eiffel). E's datahiding
works on the module-level, which can model class-level datahiding,
but enables more intelligent schemes also.

PRIVATE and PUBLIC let you declare a section of an object as visible
to the outer world or not; the outer world here is all code outside
the module. for the code within a module, everything is always visible.
example:


OBJECT mydata PRIVATE                   -> whole object is private
  bla:PTR TO mydata, burp, grrr:INT
ENDOBJECT

OBJECT aaargh
  blerk:PTR TO aaargh                   -> public
PRIVATE
  x:INT, y:INT, z:INT                   -> private
PUBLIC
  hmpf[10]:ARRAY OF mydata              -> public again
ENDOBJECT


an object is by default public, an occurring PRIVATE or PUBLIC
acts as a toggle-switch to the objects current visibility. In the
first object, all is private. The second object has only (x,y,z)
as private. PRIVATE and PUBLIC keywords may occur:
- in the object header line
- as a line on itself in the object-def
- preceding decls in an object-def
(i.e virtually anywhere)

Why datahiding?
If you want to know why datahiding is a good technique, you'd
probably want to read a good book on OO. But in short: it is
generally assumed that lots of problems in maintaining and
enhancing large pieces of software is the fact that it's hard to change
things because lots of code start to depend on certain structures
in your program. if you datahide an object, only the code within a
module will rely on the format of objects, and you can easily change
both representation of an object (for example changing a stack
implementation from ARRAY to a linked list) and the code that
works with it. If a lot of code of a large app depend on the fact
that the stack is an ARRAY, you won't be able to simply change
it, which will lead to kludges. In general, try to datahide as
much as possible without becoming too restrictive on the use
of your object. Using methods (below) will often enable you to
keep the whole object private.


@endnode
@node CH_14D
@title "14D. methods and virtual methods"
14D. methods and virtual methods
--------------------------------
A method is much like a PROC, only now it's part of an OBJECT. It
also allows you to exploit Polymorhism on objects, as we'll see
below. definition of a method:


OBJECT blerk PRIVATE
  x:PTR TO blerk, y:INT, z
ENDOBJECT

PROC getx() OF blerk IS self.x


the 'OF blerk' part tells the compiler it belongs to the object
'blerk'. Note that apart from the 'OF' the syntax is completely
like a 'PROC', however, it can't be invoked as one, and also
functions quite differently.

'self' is a local variable that is available in every method, and
is a pointer to the object that the method belongs to (in this
case 'self:PTR TO blerk'). This function just returns the value
of the x field of blerk, which actually makes sense, given that
this allows you to later change whatever x represents.

we may call methods similar to object selections ".":


DEF a:PTR TO blerk
NEW a
...
a.getx()           -> invoke method getx() on object a


in this example, upon invocation 'a' becomes the value of 'self'
during the execution of getx().

so far the use of methods has been nice, but hasn't show us its
real power, which only comes when used with inheritance.

If I inherit an object that has methods, I automatically get
those in the new object:


OBJECT burp OF blerk PRIVATE        -> same as blerk, + extra field
  prut:INT
ENDOBJECT

DEF b:PTR TO burp
NEW b
...
b.getx()                            -> same method


The interesting thing is now, that instead of inheriting a method,
you may also redefine it:


PROC getx() OF burp IS self.x+1


(it goes without saying that we may also add new methods)
so where appropriate, we can choose to modify slightly the
behaviour of methods we get from other objects, while the
interface to it (i.e. 'getx()') stays the same. Not only
does this allow us to reuse code selectively, we can also make use
of polymorhism:


PROC dosomething(o:PTR TO blerk)
  ...
  o.getx()
  ...
ENDPROC

dosomething(a)
dosomething(b)


we may call that PROC with both a and b, since both are compatible
with a blerk object. But which of the two method implementations of
getx() is invoked at o.getx()? Answer: both are. Method calls in E
are what virtual method calls are in other languages: they dynamically
act on the real type of an object (o) and call the appropriate method.

A more clear example:


-> classical OO polymorphic example

OBJECT loc
  PRIVATE xpos:INT, ypos:INT
ENDOBJECT

OBJECT point OF loc
  PRIVATE colour:INT
ENDOBJECT

OBJECT circle OF point
  PRIVATE radius:INT
ENDOBJECT

PROC show() OF loc IS WriteF('I''m a Location!\\n')
PROC show() OF point IS WriteF('I''m a Point!\\n')
PROC show() OF circle IS WriteF('I''m a Circle!\\n')

PROC main()
  DEF x:PTR TO loc,l:PTR TO loc,p:PTR TO point,c:PTR TO circle
  ForAll({x},[NEW l,NEW p,NEW c],`x.show())
ENDPROC


In the above, x is a PTR TO loc, so many would expect x.show() to
write 'I'm a Location' three times, but instead it writes the
right string for each object.

If one would write this example in a non-OO language, one would need
a SELECT for every operation like show(), testing some value
present in the object to see what it is. If I would add a new shape
to this, say:

OBJECT ellipse OF circle

I would need to change all SELECTs throughout my app to account for
it. With object-polymorhism, I can just write a show() method, and
ALL code throughout my app that calls x.show() will act correctly
when x is a PTR TO ellipse, even without recompilation!


It's difficult to show why this is powerful in a few examples, and
best to discover this is using it in real life apps. and: like I said,
read a book on it.


How does polymorphism work?
In the above examples, it's clear the compiler can't know
what method it's going to call. that's why E uses a 'class
object', and every object created gets a ptr to this object.
In the class object, all information is stored that is common
for all objects of that type, such as pointers to methods.
when the E compiler sees a call like x.show(), instead of
looking directly at the show() that belongs to the type of
x (i.e. loc), it will generate code to retrieve the pointer
to the show() method from loc's class object. since the
class object for point looks the same as loc (only maybe
slightly larger), that code will automatically call point's
show(), when x is really a point object. This is sometimes
called runtime binding.

Objects that have methods therefore always are 4 bytes larger
than you expect them to be, since they contain a class object
pointer. This pointer is automatically installed by NEW, which
is the reason _currently_ NEW is the only way to create such
an object (!).

If a method is declared with the sole purpose of enabling
subclasses to redefine it (this type of class is known as
a virtual baseclass in some languages), one may use EMPTY:

PROC bla() OF obj IS EMPTY

it may then be redefined in subclasses. Since a programmer
might not implement all methods at once, it is not an error
when the above method is executed. it will just return 0 or NIL.


One may effectively add methods to system OBJECTs:

        OBJECT mygadget OF gadget       -> from intuition!
          -> extra fields here
        ENDOBJECT

        PROC creategadget() OF mygadget IS ...

A pointer to an object such as mygadget above is then compatible
with a normal gadget pointer, i.e. it may be added directly to a
window etc.


@endnode
@node CH_14E
@title "14E. Constructors, Destructors and Super-Methods"
14E. Constructors, Destructors and Super-Methods
------------------------------------------------

The constructor name may be anything, but it is usually given the same name
as the class. One may even have multiple constructors for one class.
A constructor is called directly on the object created with NEW (and only
NEW!):

        NEW obj.stack()

Destructors however have to be named "end". An object is destroyed like
this:

        END obj

If 'obj' has a .end() method, it is automatically called. end() shouldn't
have any arguments, and it is of no use returning a value.
(see @{" 5M " link CH_5M } for a desciption of END's precise functioning).

The super-method of a method is the method by the same name of its super
class. Sometimes it's handy to call this method because you might want to
add its behaviour to the implementation of your class, however, since
you're redefined it, calling the super-method by that name will just
call yourself (!). The SUPER keyword allows you to call any method of your
superclass (or someone else's superclass):

        SUPER obj.method()

This piece of code above can be used as expression and statement.
Care has to be taken though, since if your supermethod calls another
method of that object, it will call the redefined version, not the
one at its own 'level', so to speak (generally this is what one wants).
Also, the compiler looks at the static type of 'obj' to find the
superclass, not the dynamic type (though it may still have that behaviour).



@endnode
@node CH_14F
@title "14F. Unions"
14F. Unions
-----------

Within an object a member will always have a single specific type.
However sometimes it is useful to be able to access one or members in
different context specific ways. So for example you may have an object
that stores data but the type of that data can vary at runtime. UNION's
allow you to define different structures within your object that can all
reference the same data. eg.

OBJECT myObject
  name:PTR TO CHAR
  priority:INT
  type:INT
  UNION
   [
   v1:CHAR
   v2:CHAR
   ]
   [
      v:INT
   ]
  ENDUNION  

With this object you can either access v1 and v2 or v depending on whether
you wish to store two CHAR values or an INT. The space reserved in the
object will only be 2 bytes and both v1 and v will point to the same memory
area.


@endnode

@node CH_15A
@title "15A. identifier sharing"
15A. identifier sharing
-----------------------
As you've probably guessed from the example in chapter 5D, assembly
instructions may be freely mixed with E code. The big secret is, that
a complete assembler has been built in to the compiler.
Apart from normal assembly addressing modes, you may use the following
identifiers from E:

mylabel:
LEA mylabel(PC),A1            /* labels */

DEF a                         /* variables */
MOVE.L (A0)+,a                /* note that <var> is <offset>(A4) (or A5) */

MOVE.L dosbase,A6             /* library call identifiers */
JSR    Output(A6)

MOVEQ  #TRUE,D0               /* constants */


E-VO's assembler supports following constructs,

where
n       = registernum
x       = index
lab     = label, from: "label:" or "PROC label()"
abs     = absolute addressing
s       = size. L, W or B where appropriate.

- addressing modes supported by E-VO:

  Dn, An, (An), (An)+, -(An), x(An), x(An,Dn.s),
  lab(PC), lab(PC,Dn.s), abs, abs.W

  (note: write abs.W in hexadecimal, to not confuse it with a float
   value, i.e. write   MOVE.L $4.W,A6   )

- supported partially:

  #<constexp>

- not supported:

  lab (same as abs), #lab
  use LEA lab(PC),An instead.

- extra modes:

  var.s (transfers contents of var. optionally size is ".W" or ".B",
         default is ".L")

  LibraryFunction(A6)

  example:

  ...
  MOVE.W myvar.W,D0     -> move lowword of 'myvar'
  ...
  MOVE.L dosbase,A6
  JSR    Write(A6)



As an extra, E allows to directly return registers from a function:

  ENDPROC D0

You may even interpret this as multiple return values, i.e. D0/D1/D2.


@endnode
@node CH_15B
@title "15B. the inline assembler compared to a macro assembler"
15B. the inline assembler compared to a macro assembler
-------------------------------------------------------
The inline assembler differs somewhat from your average macro-assembler,
and this is caused mainly by the fact that it is an extension to E,
and thus it obeys E-syntax. Main differences:

- comments are with /* */ and not with ";", they have a different meaning.
- keywords and registers are in uppercase, everything is case sensitive
- no macros and other luxury assembler stuff (well, there's the complete
  E language to make up for that ...)
- You should be aware that registers A4/A5 may not be trashed by inline
  assembly code, as these are used by E code. Also, if your code
  can be called by code that is register-allocated, you should preserve
  D3-D7. an instruction like

        MOVEM.L D3-D7,-(A7); /* inline asm */; MOVEM.L (A7)+,D3-D7

  should help if problems occur.
- no support for LARGE model/reloc-hunks in assembly _YET_.
  This means practically that you have to use (PC)-relative addressing
  for now (which is faster anyway).


@endnode
@node CH_15C
@title "15C. ways using binary data (INCBIN/CHAR..)"
15C. ways using binary data (INCBIN/CHAR..)
-------------------------------------------

INCBIN

syntax:         INCBIN <filename>

includes a binary file at the exact spot of the statement, should
therefore be separate from the code. Example:

mytab: INCBIN 'df1:data/blabla.bin'

LONG, INT, CHAR

syntax:         LONG <values>,...
                INT <values>,...
                CHAR <values>,...

Allows you to place binary data directly in your program. Functions much
like DC.x in assembly. Note that the CHAR statement also takes strings,
and will always be aligned to an even word-boundary. Example:

mydata: LONG 1,2; CHAR 3,4,'hi folks!',0,1


INCBIN, LONG and in general inline assembly can also be written
between PROCs in an E source (so it doesn't stand in the way of normal
E control flow). Notice however that an E source needs atleast
one PROC before any inline assembly.


@endnode
@node CH_15D
@title "15D. OPT ASM"
15D. OPT ASM
------------
OPT ASM is discussed also in chapter 16A. It allows you to operate
'E-VO' as an assembler. There's no good reason to use E-VO over some
macro-assembler, except that it is significantly faster than for example
A68k, equals DevPac and loses of AsmOne (sob 8-{). You will also have
a hard time trying to squeeze your disks of old seka-sources through E-VO,
because of the differences as described in chapter 15B. If you want to write
assembly programs with E-VO, and want to keep your sources compatible with
other assemblers, simply precede all E-specific elements with a ";",
E-VO will use them, and any other assembler will see them as a comment.
Start all regular comments with a smiley (";->").
Example:

; OPT ASM

start:  MOVEQ   #1,D0           ;-> do something silly
        RTS                     ;-> and exit

this will be assembled by any assembler, including E-VO


@endnode
@node CH_15E
@title "15E. Inline asm and register variables"
15E. Inline asm and register variables
---------------------------------------
register variables are a great companion to inline assembly, as
they function just as registers, but at the same time have clear
identifiers instead of Dx, and also are automatically saved and
restored by E code. example:

PROC bla()
  DEF count:REG
  MOVEQ #10,count
loop: WriteF('count=\\d\\n',count)
  DBRA count,loop
ENDPROC

all instruction that can work with a Dx EA, work with register
variables. examples:

MOVEQ #1,a
MOVEM.L D0/D1/a/b/A0,-(A7)
LSL.L a,b

etc.

as may be known, E-VO uses D3-D7 for these register variables. If you wish
to write code that freely mixes assembly with E, it's advisable to
keep longer-term values in register variables, and temporaries in
D0-D2/A0-A3/A6


@endnode
@node CH_16A
@title "16A. the OPT keyword"
16A. the OPT keyword
--------------------

OPT, LARGE, STACK, ASM, NOWARN, DIR, OSVERSION, MODULE, EXPORT, RTD, REG

syntax:         OPT <options>,...

allows you to change some compiler settings:

LARGE           Sets code and data model to large. Default is small;
                the compiler generates mostly pc-relative code, with a
                max-size of 32k. With LARGE, there are no such limits,
                and reloc-hunks are generated (see @{" 0C " link CH_0C }, LARGE).
STACK=x         Set stacksize to x bytes yourself. Only if you know what
                you are doing. Normally the compiler makes a very good
                guess itself at the required stack space (see @{" 16C " link CH_16C }).
ASM             Set the compiler to assembly mode. From there on, only
                assembly instructions are allowed, and no initialisation
                code is generated. (see @{" 15D " link CH_15D }, inline assembly)
NOWARN          Shut down warnings. The compiler will warn you if it
                *thinks* your program is incorrect, but still syntactically
                ok. (see @{" 0C " link CH_0C }, -n)
DIR=moduledir   Sets the directory where the compiler searches for modules.
                default='emodules:'
OSVERSION=vers  Default=33 (v1.2). Sets the minimum version of the kickstart
                (like 37 for v2.04) your program runs on. That way, your
                program simply fails while the dos.library is being opened
                in the initialisation code when running on an older machine.
                However, checking the version yourself and giving an
                appropriate error-message is more helpful for the user.
MODULE          denotes this source to be a module. (see @{" 10C " link CH_10C })
EXPORT          automatically export all declarations in a module
RTD             generates RTD's instead of RTS in the main source.
                020+ only. [experimental optimisation]
020,881,882,040,060
                generate code for these CPUs. Currently only opmises for 020
                so fpu and higher CPU's have no additional effect (except in
                the case of FPEXP below).
REG=n           use n register for register-allocation.
RUNBG           Compiles the program with a modified startup code that
                detaches from the current shell process
PURE            Checks to make sure the code generated is pure (re-entrant)
                so no use of variables in immedate lists is allowed and no
                use of GetA4(). Doing either of these will result in an
                error.
                Library code should be pure since it may be held in memory
                and called from multiple applications at the same time.
STRMERGE        Add an extra step into the compilation process where all
                constant strings values are de-duped.
POOL(memtype, puddlesize, threshsize)
                create pool that can be used later in
                your program using Alloc() and Free(). Parameters are
                optional and default to $10001 (MEMF_CLEAR|MEMF_PUBLIC),
                10240 and 256. When using OPT POOL memory for e-lists, objects
                and other memory functions will be allocated from the POOL.
                Pools are only availablein OS v39 and above.
FPEXP           Tells the compiler to generate FPU based floating point math
                code. You also need to specify either the cpu like 040/060
                or the fpu like 881/882 to switch this on.
INLINE          The compiler will create inline code for many of the smaller
                E-VO functions each time they are used instaad of calling a
                subroutine each time. This results in larger code that will
                run slightly faster.
NOSTARTUP       Disables as much of the startup code as possible. No
                libraries will be automatically opened.                
UTILLIB         Automatically opens UTILITY.LIBRARY in addition to the
                usual libraries that E-VO will open. The library pointer
                will be placed in UTILBASE
LEGACYINT       The Int() function did not sign extend the value. This is
                not in line with the definition of INT variable type which
                is signed. This has been corrected in v3.6.0 of E-VO and
                this option will revert that function back to the previous
                way of working for backwards compatibility.

example:

OPT STACK=20000,NOWARN,DIR='df1:modules',OSVERSION=39,REG=3


@endnode
@node CH_16B
@title "16B. small/large model"
16B. small/large model
----------------------
Amiga E lets you choose between SMALL and LARGE code/data model.
Note that most of the programs you'll write (especially if you just
started with E) will fit into 32k when compiled: you won't have to
bother setting some code-generation model. You'll recognise the
need for LARGE model as soon as E-VO starts complaining that it can't
squeeze your code into 32k anymore. To compile a source with LARGE model:

1> evo large sizy.e

or better yet, put the statement

OPT LARGE

at the top of your code.


@endnode
@node CH_16C
@title "16C. stack organisation"
16C. stack organisation
-----------------------
To store all local and global variables, the run-time system of an
executable generated by Amiga E allocates a chunk of memory,
from which it takes some fixed part to store all global variables.
The rest will be dynamically used as functions get called.
as a function is called in E, space on the stack is reserved
to store all local data, which is released upon exit of the function.
That is why having large arrays of local data can be dangerous when
used recursively: all data of previous calls to the same function
still resides on the stack and eats up large parts of the free stack
space. However, if PROCs are called in a linear fashion, there's
no way the stack will overflow.
Example:

global data:            10k (arrays etc.)
local data PROC #1:      1k
local data PROC #1:      3k

the runtime system always reserves an extra 10k over this for normal
recursion (for example with small local-arrays) and additional buffers/
system spaces, thus will allocate a total of 24k stack space


@endnode
@node CH_16D
@title "16D. hardcoded limits"
16D. hardcoded limits
---------------------

Note these signs: (+-)    just about, depends on situation,
                  (n.l.)  no clear limit, but this seems reasonable.

--------------------------------------------------------------------------
OBJECT/ITEM                                     SIZE/AMOUNT/MAX
--------------------------------------------------------------------------

value datatype BYTE                             -128 .. 127
value datatype CHAR                             0 .. 255
value datatype INT                              -32768 .. +32767
value datatype WORD                             0 .. 65535
value datatype LONG/PTR                         -2 gig .. +2 gig

identifierlength                                100 bytes (n.l.)
length of one source line                       2000 lexical tokens (+-)
source length                                   2 gig (theoretically)
constant lists                                  few hundred elements (+-)
constant strings                                1000 chars (n.l.)
max. nesting depth of loops (IF, FOR etc.)      500 deep
max. nesting depth of comments                  infinite

#of local variables per procedure               8000
#of global variables                            7500
#of arguments to own functions                  8000 (together with locals)
#of arguments to E-varargs functions (WriteF()) 64 (v2.1) / 1024 (v2.5)
max length of string processed by WriteF() etc. 5000 (depends on stack)

one object (allocated local/global or dyn.)     8 k
one array, list or string (local or global)     32 k
one string (dynamically)                        32 k
one list (dynamically)                          128 k
one array (dynamically)                         2 gig
objects with NEW                                64k elem.
CHAR/INT/LONG with NEW                          2 gig.

local data per procedure                        250 meg
global data                                     250 meg

code size of one procedure                      32 k
code size of executable                         32 k SMALL, 2 gb LARGE model
current practical limit (may extend in future)  2-5 mb (v2.1) / 10 mb (v2.5)

buffersize of generated code and identifiers    relative to source
buffersize of labels/branches and intermediate  independently (re)allocated



@endnode
@node CH_16E
@title "16E. error messages, warnings and the unreferenced check"
16E. error messages, warnings and the unreferenced check
--------------------------------------------------------
Sometimes, when compiling your source with E-VO, you get a message
of the sort UNREFERENCED: <ident>, <ident>, ...
This is the case when you have declared variables, functions or labels,
but did not use them. This is an extra service rendered to you by the
compiler to help you find out about those hard to find errors.

There are several warnings that the compiler issues to notify you that
something might be wrong, but is not really an error.


- "A4/A5 used in inline assembly"
  This is the warning you'll get if you use registers A4 or A5 in your
  assembly code. The reason for this is that those registers are used
  internally by E to address the global and local variables respectively.
  Of course there might be a good reason to use these, like doing
  a MOVEM.L A4/A5,-(A7) before a large part of inline assembly code

- "keep an eye on your stacksize"
- "stack is definitely too small"
  Both these may be issued when you use OPT STACK=<size>. The compiler
  will simply match your <size> against its own estimate (see @{" 16C " link CH_16C }),
  and issue the former warning if it thinks it's ok but a bit on the small
  side, and the latter if it's probably too small.

- 'suspicious use of "=" in void expressions (s). (line %d)'
  This warning is issued if you write expressions like 'a=1' as a
  statement. One reason for this is the fact that a comparison doesn't
  make much sense as a statement, but the main reason is that it could be
  an often occurring typo for 'a:=1'. Forgetting those ":" may be hard to
  find, and it may have disastrous consequences.

- 'module changed OPT settings'
  If you use a module that has OPT OSVERSION=37, this changes the
  OPT for the main program too. this warning serves to make you aware
  of this. put such an OPT in the main program too to get rid of it.

- 'variable used as function'
  in v3, arbitrary variables may be used as function. this function
  is there to warn you so you don't accidentally do this.

- 'code outside PROCs'
  You wrote E code in between PROCs, which is only rarely useful.

- 'REG var used after EXCEPT'

- 'member used as method'
  similar to variable used as function warning but for members/methods

- 'OPT PURE recommended for libraries'
  suggestion when compiling a library that OPT PURE should be used to
  enable checking for E language features that generate non-pure code.

- 'assign of ARRAY/STRING/LIST variable.'
  if a variable is declared as an ARRAY/STRING/LIST then assigning to
  this variable may well not be the intended action.


Errors.
The compiler will print the source-code-line that caused the error below it,
with a cursor at the exact spot.  The cursor denotes the spot the compiler
was when it _discovered_ the error, it is thus likely that the token that
caused the error is the one or two tokens just _before_ the cursor, maybe
even on the previous line (i.e. the cursor is always _after_ the error).

- 'syntax error'
  Most common error. This error is issued either when no other
  error is appropriate or your way of ordering code in your sources
  is too abnormal.

- 'unknown keyword/const'
  You have used an identifier in uppercase (like "IF" or "TRUE"), and
  the compiler could not find a definition for it. Causes:
  * mispelled keyword
  * you used a constant, but forgot to define it in a CONST statement
  * you forgot to specify the module where your constant is defined

- '":=" expected'
  You have written a FOR statement or an assignment, and put something
  other than ":=" in its place.

- 'unexpected characters in line'
  You used characters that have no syntactic meaning in E outside of
  a string. examples: "@!&\\~"

- 'label expected'
  At some places, for example after the PROC or JUMP keyword,
  a label identifier is required. You wrote something else.

- '"," expected'
  In specifying a list of items (for example a parameter list)
  you wrote something else instead of a comma.

- 'variable expected'
  This construction requires a variable, example:
  FOR <var>:= ... etc.

- 'value does not fit in 32 bit'
  In specifying a constant value (see @{" 2A " link CH_2A }-2E) you wrote too
  large a number, examples:  $FFFFFFFFF, "abcdef".
  Also occurs when you define a SET of more than 32 elements.

- 'missing apostrophe/quote'
  You forgot the ' at the other end of a string.

- 'incoherent program structure'
  * you started a new PROC before ending the last one
  * you don't nest your loops properly, for example:
    FOR
      IF
      ENDFOR
    ENDIF

- 'illegal command-line option'
  In specifying 'EVO -opt source' you wrote something for '-opt'
  that is not a legal option to E-VO.

- 'division and multiplication 16bit only'
  The compiler detected that you were about to use 32bits
  for * or /. This would not have the desired result at runtime.
  (see @{" 9G " link CH_9G }, Mul() and Div()).

- 'superfluous items in expression/statement'
  After the compiler already compiled your statement, it still found
  lexical tokens instead of an end of line. You probably forgot
  the <lf> or ";" to separate two statements.

- 'procedure "main" not available'
  Your program does not include a main procedure !

- 'double declaration of label'
  You declared a label twice, for example:
  label:
  PROC label()

- 'unsafe use of "*" or "/"'
  This again has to do with 16bit instead of 32bit * and /.
  See 'division and multiplication 16bit only'.

- "reading sourcefile didn't succeed"
  Check your source spec. that you gave with 'evo mysource'
  make sure the file ends in '.e'

- "writing executable didn't succeed"
  Trying to write the generated code as an executable caused a dos
  error. For example, the executable that did already exist could
  not be overwritten.

- 'no args'
  "USAGE: evo [-opts] <sourcecodefilename> ('.e' is added)"
  You get this by just typing 'evo' without any arguments.

- 'unknown/illegal addressing mode'
  This error is reported only by the inline assembler. Possible causes are:
  * you used some addressing mode that does not exist on the 68000
  * the addressing mode exists, but not for this instruction.
    not all assembly instructions support all combinations of
    effective addresses for source and destination.

- 'unmatched parentheses'
  Your statement has more "(" than ")" or the other way around

- 'double declaration'
  One identifier is used in two or more declarations.

- 'unknown identifier'
  An identifier is not used in any declaration; it is unknown.
  You probably forgot to put it in a DEF statement.

- 'incorrect #of args or use of ()'
  * You forgot to put "(" or ")" at the right spot
  * you supplied the incorrect #of arguments to some function

- 'unknown e/library function'
  You wrote an identifier with the first character in uppercase, and
  the second in lowercase, but the compiler could not find a definition.
  Possible causes:
  * Misspelled name of function
  * You forgot to include the module that defines this library call.

- 'illegal function call'
  Rarely occurs. You get this one if you try to construct weird
  function calls like nested WriteF()'s. Example:
  WriteF(WriteF('hi!'))

- 'unknown format code following "\\"'
  You specified a format code in a string which is illegal.
  (see @{" 2F " link CH_2F } for a listing of format codes)

- '/* not properly nested comment structure */'
  The #of '/*' is unequal to the #of '*/', or is placed in a funny order.

- 'could not load binary'
  <filespec> in INCBIN <filespec> could not be read.

- '"}" expected'
  You started an expression with "{<var>" , but forgot the "}"

- 'immediate value expected'
  Some constructions require an immediate value instead of an expression.
  Example:
  DEF s[x*y]:STRING /* wrong: only something like s[100]:STRING is legal */

- 'incorrect size of value'
  You specified an unacceptably large (or small) value for
  some construction.

  Examples:
  DEF s[-1]:STRING, t[1000000]:STRING    /* needs to be 0..32000  */
  MOVEQ #1000,D2                         /* needs to be -128..127 */

- 'no e code allowed in assembly modus'
  You wish to operate the compiler as an assembler by writing 'OPT ASM',
  but, by accident, wrote some E code.

- 'illegal/inappropriate type'
  At someplace where a <type> spec. was needed, you wrote something
  inappropriate. Examples:
  DEF a:PTR TO ARRAY       /* no such type */
  [1,2,3]:STRING

- '"]" expected'
  You started with "[", but never ended with "]"

- 'statement out of local/global scope'
  A breakpoint of scope is the first PROC statement. before that,
  only global definitions (DEF,CONST,MODULE etc.) are allowed, and no code.
  In the second part, only code and function definitions are legal, no
  global definitions.

- 'could not read module correctly'
  A dos error occurred while trying to read a module from a MODULE
  statement. Causes:
  * emodules: was not assigned properly
  * module name was misspelled, or did not exist in the first place
  * you wrote MODULE 'bla.m' instead of MODULE 'bla'

- 'workspace full!'
  Rarely occurs. If it does, you'll need the '-m' (ADDBUF) option to
  manually force E-VO to make a bigger estimate on the needed amount of
  memory. Try compiling with -m2, then -m3 until the error disappears.
  You'll probably be writing huge applications with giant amounts
  of data just to even possibly get this error.

- 'not enough memory while (re-)allocating'
  Just like that. Possible solutions:
  1. You were running other programs in multitasking. Leave them and try
    again.
  2. You were low on memory anyway and your memory was fragmented.
     Try rebooting.
  3. None of 1-2. Buy a memory expansion (um).

- 'incorrect object definition'
  You were being silly while writing the definitions between OBJECT and
  ENDOBJECT. (see @{" 8F " link CH_8F } to find out how to do it right).

- 'illegal use of/reference to object'
  If you use expressions like ptr.member, member needs to be a legal
  member of the object ptr is pointing to.

- 'incomplete if-then-else expression'
  If you use IF as an operator (see @{" 4E " link CH_4E }), then an ELSE part
  needs to be present: an expression with an IF in it always needs to
  return a value, while a statement with an IF in it can just 'do nothing'
  if no ELSE part is present.

- 'unknown object identifier'
  You used an identifier that was recognised by the compiler as being
  part of some object, but you forgot to declare it. Causes:
  * misspelled name
  * missing module
  * the identifier in the module is spelled not like you expected
    from the RKRM's. Check with ShowModule.
    Note that amiga-system-objects inherit from assembly identifiers,
    not from C. Second: identifiers obey E-syntax.

- 'double declaration of object identifier'
  One identifier used in two object definitions

- 'reference(s) out of 32k range: switch to LARGE model'
  Your program is growing larger than 32k. Simply put 'OPT LARGE'
  in your source and code on. (see @{" 16B " link CH_16B }).

- 'reference(s) out of 256 byte range'
  You probably wrote BRA.S or Bcc.S over too great a distance.

- 'too sizy expression'
  You used a string '' or list [], possibly recursive [[]], that is too
  sizy.

- 'incomplete exception handler definition'
  You probably used EXCEPT without HANDLE, or the other way round
  (see @{" 13A " link CH_13A } on exception handling).

- 'not allowed in a module'
  You're doing one of the few things you can't do in a module,
  such as global variables with initialisations.

- 'allowed in modules only'
  you probably use EXPORT in your main source

- 'this doesn't make sense'
  general error.

- 'you need a newer version of E-VO for this :-)'
  You probably use a module that was compiled with a newer
  version of E-VO than you currently have.

- 'no matching "["'
  within a statement, a "]" was found, without a matching "]".

- 'this instruction needs a better CPU/FPU (see
  You use a construction (probably an asm instruction) that
  requires an OPT 020 or the like.

- 'object doesn't understand this method'
  you invoke a method on a object that wasn't defined
  for its type.

- 'method doesn't have same #of args as method of baseclass'
  If you redefine a method, you have to make sure the new one
  has the same #of args as the original.

- 'too many register variables in this function'
  If you use :REG to assign register variables yourself,
  you can't use more than 5, currently.

- 'Linker can't find all symbols'
  If you use a module A that uses again a module B,
  B also needs to linked. A relies on certain PROCs to
  be present in B, and if B was recompiled with those
  PROCs removed, the linker has trouble putting your
  exe together.

- 'could not open "mathieeesingbas.library"'
  If you use float code, the compiler itself may need
  float functions to be able to generate code.

- 'illegal destructor definition'
  You defined an end() method with arguments (or with a returnvalue)

- 'implicit initialisation of private members'
  You write a [...] or NEW [...] expression that has private parts.

- 'double method declaration'
  You defined a method on this object twice.

- 'object referenced by other object not found'
  You probably inherited from some other object in another module, and
  then changed it (its name for example) without changing/recompiling
  other dependant modules too.

- 'unknown preprocessor keyword'
  only #define, #ifdef, #ifndef and #endif are known by E-VO's PP.

- 'illegal macro definition'
  You made a syntax error in typing your #define (see @{" 17L " link CH_17L })

- 'incoherent #ifdef/#ifndef nesting'
  You forgot to close with #endif or similar

- 'macro redefinition'
  You can't use the same macro-name-identifier twice.

- 'syntax error in #ifdef/#ifndef/#else/#endif'
  (see @{" 17L " link CH_17L } for the correct way to write conditionally compiled code)

- 'macro(s) nested too deep'
  You will get this if macros expand to other macros which expand to yet
  others... such that the amount of memory needed for this is getting
  out of hand. More likely you just defined a recursive macro (which will
  want to expand forever).

- 'method definition out of object/module scope'
  You can only define methods for an object in the same module/source
  where that object is defined.

- 'library definition problem'
  Generally when there's something wrong with the procs you specified in
  the LIBRARY declaration, i.e.: one doesn't exist, is not a PROC or is
  a method etc.

- 'object not known at this point'
  You are calling a method on an object whose type is not fully known
  at that point in the source. Better place the OBJECT at the top of
  your source.


@endnode
@node CH_16F
@title "16F. compiler buffer organisation and allocation"
16F. compiler buffer organisation and allocation
------------------------------------------------
When you get the error 'workspace full' (very unlikely), or want
to know what really happens when your program is compiled, it's useful
to know how E-VO organizes its buffers.

A compiler, and in this case E-VO needs buffers to keep track of all sorts
of things, like identifiers etc., and it needs a buffer to keep the
generated code in. E-VO doesn't know how big these buffers need to be.
for most buffers, like the one for various strcutures, this is no
problem: if the buffer is full while compiling, E-VO just allocates a
new piece of memory and continues. Other buffers, like the one for
the generated code, need to be a continuous block of memory that doesn't
move while compiling: E-VO needs to make a pretty good estimate of
this buffersize to be able to compile small and large sources alike.
To do this, E-VO computes the needed memory relative to the size of
your source code, and adds a nice amount to it. This way, in 99% of the
cases, E-VO will have allocated enough memory to compile just about any
source, in other cases, you'll get the error and have to specify more
memory with the '-m' (ADDBUF) option.

Experiment with different types and sizes of example-sources in combination
with the '-b' (SHOWBUF) option (see @{" 0C " link CH_0C }) to see how this works in practice.


@endnode
@node CH_16G
@title "16G. register allocation"
16G. register allocation
------------------------
E v3 supports a register allocation, which is a technique to keep
variables in registers instead of on the stack. For normal code
that uses OS-routines you won't notice the difference very much,
but for tight computation-loops, this optimisation can make a big
difference. There are two ways to use register allocation:

- with the option REG.
  If you write for example EVO REG=3 bla.e, (max=5, currently),
  E-VO will compute for each PROC the 3 most-used variables in
  registers. Register allocation is a technique that tries to be
  intelligent: it will compute for each var a weight, and will
  use heuristics to increase that weight, for example a var used
  in a FOR loop gets relatively a higher weight than one outside
  it, and one in an IF gets an even lower weight. These weights
  are combined, so a WHILE in a FOR gets quite a high weight.

- DIY: you can put the keyword REG in front of any type in a
  declaration, for example:

        DEF x:REG, s[4]:REG LIST

  you can do this if you don't trust the register-allocator,
  or if you want to fine-tune just one PROC. You can even use
  both together: if in a PROC you have one var with :REG,
  compiling with REG=5 will allow E-VO to pick the remaining 4
  by itself.

The default is REG=0, so E-VO works much like the older versions.

The variables that CAN be allocated are only local variables
that are not parameters. also, if you take the address of
a variable with {} it can't be put in a register either (guess
why...). registers can't be allocated in PROCs that have
an exception handler, for now.

There are a few things to note when using registers:
- this part of E-VO is currently was tested to be pretty reliable,
  but you still check that behaviour is  the same as in non-allocated code.
- E-VO uses registers D7..D3 for variables, so if you use
  inline assembly, you need to check that PROCs that use
  register-allocation or :REG don't trash these (see @{" 15E " link CH_15E }).
  The code generated for a PROC automatically saves the
  registers it uses (callee save) to protect the code that
  called it.
- hint: compiling with REG=5 is not inherently fastest,
  since variable saving on function/library calls also
  incurs an overhead. REG=3 may be better for some
  cases. Also if _all_ code in question deals with
  library calls instead of pure computation, expect no
  gain from registers.

-> register allocation will easily make this program twice as fast

PROC main()
  DEF a,b=10,c=20,d
  FOR a:=1 TO 1000000 DO d:=b+c
ENDPROC

-> at most 5% faster when using register allocation

PROC main()
  DEF a,s[100]:STRING,t
  t:='putting "a" in a reg won''t give that much of a speedup, I think.'
  FOR a:=1 TO 100000 DO StrCopy(s,t)
ENDPROC



@endnode
@node CH_17A
@title "17A. bin/showmodule"
17A. bin/showmodule
-------------------
As you might have noticed, E's equivalent for "includes", modules, are
binary files, much like those usually suplied with Modula2 compilers, for
example. To display the contents of such a file in readable ascii form, you
may use

showmodule:

showmodule <modulespec>

examples:

1> showmodule emodules:intuition/intuition
1> showmodule >gadtools.txt emodules:gadtools

note that showmodule by default outputs to stdout, and may be interrupted at
any point by <ctrlc>.

when given the option -c, ShowModule will output modules with library
declarations in .fd format, constants in #define format.

The -e option will make ShowModule format the output as an E file that
can be recompiled.

@endnode
@node CH_17B
@title "17B. sources/utilities/showhunk.e, bin/showhunk"
17B. sources/utilities/showhunk.e, bin/showhunk
-----------------------------------------------
Displays all types of executable files, and also object ".o" files as
generated by (other) compilers/assemblers.  will show you the (very simple)
structure of executables generated by E-VO, but also supports complex
overlay-files.  also dumps labels (like XREF's and XDEF's).

most important of all, ShowHunk features a disassembler for
code-hunks. use the option 'DISASM/S'

showhunk <exefile>

like:   1> showhunk helloworld
        1> showhunk disasm dpaint


@endnode
@node CH_17C
@title "17C. bin/iconvert, bin/pragma2module"
17C. bin/iconvert, bin/pragma2module
------------------------------------
These two utilities are for advanced E-programmers only. if
you don't feel like one (yet), skip this part.

[NOTE: like the showmodule utility, the sources to these utilities
 have been removed from the distribution, because people misused
 their knowledge of the .m module format. It is PRIVATE. Contact
 me first if you want to do something with it.]

Iconvert will convert structure and constant definitions in
assembly ".i" files to E modules, and pragma2module will do the
same for SAS/C pragma library definition files. Of course, all
commodores includes have already been converted this way, but
say you find a nice PD library that you may want to use with E,
you will need these utilities.

most libraries come with various includes defining, most obvious,
the library calls of the library, as well as the constants
and structures (OBJECTs in E) that it uses. say that it is
called "tools.library", then it will probably feature:

pragmas/tools_pragmas.h
includes/tools.i

then do:

1> pragma2module tools_pragmas.h

rename the resulting "tools_pragmas.m" to "tools.m" and put it
in emodules:, check with ShowModule if all went well.

Now, in your program you may use tools.library:

MODULE 'tools'

PROC main()
  IF (toolsbase:=Openlibrary('tools.library',37))=NIL THEN error()

  ToolsFunc()

...etc.

convert tools.i with Iconvert to another tools.m, which you place in
emodules:libraries, for example. Iconvert needs an assembler like
the PD A68k or PhxAss (use option "-p") to do the hard work of understanding
the actual assembly.

1> iconvert tools.i

see with showmodule what became of the ".i" file. use in your
program with:

MODULE 'libraries/tools'

DEF x:toolsobj, y=TOOLS_CONST

converting with Iconvert may require some assembly expertise, as
Iconvert relies on the correct format of the ".i" file, just like
commodores assembly includes. About 10% of the ".i" files need
to be patched by hand to be "convertable". definitions that
Iconvert judges correctly are amongst others

<label> EQU <any_expression>

STRUCTURE <sname>,0      ; if <>0, then   <struct>_SIZEOF
ULONG <sname>_<label>
BPTR <sname>_<label>
   ; etc.
LABEL <sname>_SIZEOF     ; or "_SIZE"

to get an idea what kind of assembly-expression Iconvert can
handle, take a look at commodores assembly includes and compare
them to the equivalent modules (for example intuition.i).


@endnode
@node CH_17D
@title "17D. bin/ShowCache, bin/FlushCache"
17D. bin/ShowCache, bin/FlushCache
----------------------------------
The E Module Cache is a piece of memory that is able to hold modules (.m)
between compiles.  The first time you use a certain module, E-VO will load
it from disk and put it in the cache.  the second time and on, E-VO will
find it in cache and doesn't have to load anything from disk.  If E-VO
compiles a module of which an old version is present in cache, it will flush
it. One can imagine that this a tremendous speedup, even for people with
HD's when they use a lot of modules and recompile often.

To see what's currently stored in the cache (and how much memory
it's wasteing :-), type:

1> ShowCache

A second utility, FlushCache, allows you to selectively remove modules
from the cache. Reasons for this can be:
- you can't afford the memory it uses
- you created a new .m, with a tool other than E-VO, so you need to flush
  by hand (in all other cases it's NOT necessary to flush!).
The argument to Flushcache is the substring that needs to occur in
a module name for it to be flushed. no args means flush all.

1> FlushCache                        ; empty whole cache
1> FlushCache intuition/             ; flush all intuition-related modules

You can use the E-VO option 'IGNORECACHE/S' to compile a source without
using the cache. Whether the cache is full or empty, E-VO will load
all from disk, and store nothing new in cache.

If two E-VO's try to access the cache simultaneously in multitasking,
the second E-VO acts as if its IGNORECACHE flag were set.


@endnode
@node CH_17E
@title "17E. rexx/ecompile.rexx"
17E. rexx/ecompile.rexx
-----------------------
This is a rexx-script for CygnusEd (tm), and enables you to compile E
programs from the editor.  Just assign this script a function key in the
editor with "Install Dos/Arexx command ..." (check your CED-manual if you're
not sure how to do this).  Now write your programs, and press Fx if you wish
to compile. Your source will be saved if necessary, the compiler will be
invoked on a separate console window, and the program is run on the same
console.  When your program is done, you may press <return> to return to the
editor (CED-screen to back and front is automatically done by the script).
If an error occurred during compilation, the script will let CED jump to the
line of error after you pressed <return>

Note:  in the script there is a path name as to where the compiler can be
found.  You probably need to change this.  Also, the script copies E-VO to
ram: for systems with a slower SYS:  device, you may want to disable this if
you have a fast HD.


@endnode
@node CH_17F
@title "17F. bin/o2m"
17F. bin/o2m
------------
If you have large pieces of assembly source that you'd like to use it would
be tedious at best to convert them all by hand to E's inline assembly.  o2m
allows you to simply have your favourite macro-assembler assemble it all to
a .o file, and o2, then will turn this .o file into a .m file for use with
E. If you have a file bla.o:

1> o2m bla

will produce bla.m.  However, the .o file will have to obey certain rules.
It should consist of just one code-hunk with external definitions (XDEFs)
for each symbol you wish to reference from E, and no XREFs.  typically,

your source would look like:

        XDEF add__ii

add__ii:
        move.l  4(a7),d0
        add.l   8(a7),d0
        rts
        
this example shows a bit of assembly code that gets two arguments (hence
the two "i" for integer). arguments can be found on the stack, where
4(a7) is the last arg, 8(a7) the one before that etc.

Showhunk shows you this:

        hunk_unit: 
HUNK -1 hunk_name:  
        hunk_code: 12 bytes
        hunk_ext
          add__ii = $0

this type of .o file is easily transformed to .m by o2m:

/* this module contains 12 bytes of code! */

PROC add(a,b)


there are a couple of things to note:
- if your asm code uses D3-D7/A4/A5 you should probably save it.
- if a label doesn't have the "__" with an "i" for each function,
  it becomes a parameterless function. Don't worry if the label
  actually references data, you can simply get the address of this
  'proc' with {}, and use it as a ptr to your data.

theoretically, o2m could be used to link C code to E programs, however in
practise this is often not feasable.  If your C compiler allows you to
'tune' the resulting .o files a bit, this might work.

some problems are:
- reference of C functions, for example _printf()
- reference of globals vars created by C startup code. C code
  may reference "DOSBase" as an XREF, whereas E's startup code makes 
  this value available somewhere on the stack
- call/register conventions.

I did manage to link a small C function to E that only does some
computation, and call it succesfully (this was done using MaxonC++, whose
linker uses the __ii convention for parameters also).


@endnode
@node CH_17G
@title "17G. bin/EYacc"
17G. bin/EYacc
--------------

This file is not included in the E-VO package.
It can be obtained in the original E compiler files from the Aminet
http://aminet.net/package/dev/e/amigae33a


This is a port of the famous Yacc utility for unix, which now produces E
code instead of C code.  It is only a first version, so don't expect too
much from it.  If you have no clue what Yacc does, read a text on it, since
I'm not going to explain that in full here.

Basically you can write .y sources as normal, only where actions used to be
written in C, now you can write E.  See the Src/Yacc/bcalc.y example.

1> eyacc bcalc.y

produces a file 'yyparse.e'

1> evo yyparse

will get you a module, which contains only the function yyparse().  The rest
of 'how to interface with Yacc' should be analoguous to C.

[note: I'm halfway through a translation of Lex to E-Lex, but not done yet.]

further info.
E-Yacc is a modification of Berkeley Yacc 1.8, originally by
corbett@berkeley.edu.  The inclusion of this modified version in the E
distribution is totally legal, as the author states in the original BYacc1.8
README:

"   Berkeley Yacc is in the public domain.  The data structures and
algorithms used in Berkeley Yacc are all either taken from documents
available to the general public or are inventions of the author.  Anyone may
freely distribute source or binary forms of Berkeley Yacc whether unchanged
or modified. Distributers may charge whatever fees they can obtain for
Berkeley Yacc. Programs generated by Berkeley Yacc may be distributed
freely. "


@endnode
@node CH_17H
@title "17H. Src/Various/SrcGen.e"
17H. Src/Various/SrcGen.e
-------------------------

This source file is not included in the E-VO package.
It can be obtained in the original E compiler files from the Aminet
http://aminet.net/package/dev/e/amigae33a

SrcGen GadToolBox source generator for E: beta version
[note: this is now very much obsolete. Have a look at EasyGUI]

You'll be needing GadToolBox v2.0 or higher, and have the
gadtoolsbox.library that comes with it in your LIBS:. Now, with GTB,
make some simple example (just a window with a few gadgets/menus etc.),
save it as "bla" (filename will be "bla.gui"), and type:

1> SrcGen bla
1> EVO bla
1> bla

"bla.e" contains the routines for opening your interface, as well
as some routines to handle idcmpmessages, errors etc., and a dummy
"main" that just waits for one selection. here you can put in
your own code. see the commandline template how to stop SrcGen from
generating these routines.

That's all there's to it. If you have problems, just check the
source that has been generated.


@endnode
@node CH_17I
@title "17I. bin/EBuild"
17I. bin/EBuild
---------------

This  file is not included in the E-VO package.
It can be obtained from the Aminet

http://aminet.net/package/dev/e/ebuild


@endnode
@node CH_17J
@title "17J. Aprof"
17J. Aprof
---------------

This file is not included in the E-VO package.
It can be obtained in the original E compiler files from the Aminet
http://aminet.net/package/dev/e/amigae33a


@endnode
@node CH_17K
@title "17K. EDBG"
17K. EDBG
---------
Updates to v3.8.0
- Added support for extra debug info generated by E-VO 3.8.0.
  Currently this extra debug info is not used by EDBG and was added to E-VO
  to support CodeCraft which has a more advanced debugger but requires
  OS 3.2 and a 68020.

Updates to v3.6.1
- Setting breakpoint caused crashes on certain versions of graphics.library

Updates to v3.6.0:
- add support for ECX/EEC compiled executables

Updates to v3.5.0:
- Added case insensitive searches
- Added search next option
- Display variables in both hex and decimal
- Multiple breakpoints
- Support for debugging while the code is running
- Add watch variables by name
- Support for debugging under OS 4.x without using CPU exceptions
- Section support

bugs fixed for v3.3a:
- Double-clicking on window frames no longer sets breakpoints, etc.
- 'ARG/K' argument now works (sets the arguments for the program being
  debugged)
- More than 16 windows allowed now, without crashing!
- Safer closing of public screen
- Unary negation now works as an expression
- No longer ignores some identifiers with "_"
- Serious (but benign?) bugs in initial breakpoint setup code removed

features new for v3.3a:
- Integrated calls to Explorer to debug objects (extracts type
  information from declarations)
- Can now debug standard global variables
- Can now debug "self" variable (requires EC v3.3a+)
- "Follow Step" mode added (and configurable)
- "Repeat Step" mode added (useful for debugging loops)
- "^var" expression now handled
- Source location of offsets added (useful for finding Enforcer hits)
- Registers now marked with whether they have changed since last step
- Status/flags register now expressed more clearly


Explorer
--------

To examine/watch a variable (in the current scope) double-click on it.
To see it in Explorer, press Shift while you double-click.  If
Explorer is not running it will be automatically started (as
"explorer" or whatever you set in the options, including any
arguments).  It will be instructed to open on the same screen as EDBG,
and will quit when EDBG quits (but only if EDBG started it).  Explorer
ought to be run in Auto-Reply mode, and you need Explorer v2.2j+.



EDBG is the E sourcelevel debugger. To use it, compile your source with
the DEBUG flag (this works on both main program an modules).
This will add debug infos to your executable/module.

NOTE: Do NOT distribute a program of which any part is compiled with
DEBUG/S. (You can check this with ShowHunk, it should not contain any
hunk_debug).
Programs with debug-infos are compiled with extra NOPs to facilitate
debugging, which isn't wanted in the final code.

Always make sure sources and compiled are in the current directory, then
fire up EDBG with:

1> EDBG exename

template:

EXECUTABLE/A,PUBSCREEN/K: 

with for example PUBSCREEN=Workbench you can run EDBG anywhere. By default
it opens it own screen, which is a clone of the workbench in size and
display mode. EDBG works fine on the graphics-cards etc.

EDBG opens a window for every source/module in your project, and it will
start with the one that contains 'main'.  From you can step through your
code, and windows will automatically open when necessary.  When code doesn't
have a source attached, it can be executed only (which is sometimes handy,
if it doesn't need to be debugged).

From here all is pretty intuitive. Major buttons are the first two pictures
which are step over/in. Step in follows the code every step as it is
executed, step over does the same but does not enter subroutines.
[Just try it, the debugger is quite intuitive]

Other functions show memory or register windows, and various other
functions. An important one is double-clicking on variable-names:  this
will show their contents in variable view window.  Double clicking
somewhere not on a variable allows you to set a breakpoint.

For the memory views, memory breakpoints, exception raising features a
subset of the E expression language can be used to denote addresses and
values:

values: 123, $ABC, %010101, "FORM"
variables: a, {a}
operators: +, -, *, /, ()

For example:
- to check when variable 'a' is modified, use '{a}' as a memory breakpoint.
- to view the memory pointed to by array 'a' (of LONGs) at offset 'b', use
  'a+(b*4)' with a memory view.
- to change the contents of variable 'a' into 'b*2' click on the variable
  in the variable-view and select 'Modify'.

Note that when raising exceptions, that the handler for the current PROC
is installed only at the first statement after the local DEFs, i.e. you will
want to step past there first before raising an exception that should
go to that handler.

To quit a debugging session in the middle of a program while still freeing
all resources, or to test the capacity of your program to deal with
failures, raise an exception.

Caveats:
- The debugged program runs on the same task as EDBG. this means that
  all code that does something special to the task will have to be
  careful. An example is Forbid() etc.
- A special case is ReadArgs(). because EDBG already read the args, a
  call from the debugged program will cause a read from the console.
  So you can conveniently type the args to your program on the EDBG
  output console and press return.

Know problems:
- The variable view treats scopes on basis of variable names

Future:
- editing in variable and memory views
- more extensive E expression language support
- E types support (OBJECTs in particular)
- cooperation with enforcer (as soon as I can run it)


The LINEDEBUG and SYM options.
The LINEDEBUG option adds linedebug info to your executable (for each line
of code inside a PROC). This option is necessary for EDBG, but the DEBUG
switch automatically turns it on. LINEDEBUG is partially compatible with the
"LINE" HUNK_DEBUG produced by other compilers/assemblers, so it can be
useful with other debug-tools as well. The SYM option is not necessary for
EDBG, but can be useful for others, such as AProf or disassemblers.


The EDBG Arexx port (name: "EDBG")

EDBG's prefs saving works by saving an arexx-script, and running it again on
startup.  It will save specific prefs for every project, remember window
positions (if you want to), and also variables you were watching!

Select 'Settings...', and after that 'Save Settings'.  You may add optional
Arexx-scripts that can be launched directly from EDBG's 'Rexx' menu.  The
saved script is '.edbg-startup.rexx' in the current dir, which you are
encouraged to examine and modify.

currently supported EDBG Arexx commands:

        QUIT

emergency exit (raising an exception is friendlier).

        STEPIN
        STEPOVER
        RUN

step in/over and run to breakpoint. only actually performed when
control returns to the debugged program.

        EVAL exp

allows you to evaluate an E expression. You may use variables
from the scope that is currently being debugged (much like
'Eval E Expression' from the menu). Nearly all Arexx commands
that expect integer arguments can use similar expressions to EVAL.
This is the only command that returns something in the rexx
variable "rc" (the value of the expression). example:

'eval {a}+(b*4)'
say 'rc =' rc

you might need to fiddle with "options failat". Note closely
how these to get their variables from entirely different scopes:

'eval a+b'
'eval' a+b

(from Arexx (!))

        BREAKPOINT linenum

sets the breakpoint to that linenumber (in the current view).

        MEMORYBREAKPOINT addr

sets a memory breakpoint on that address. example:

'memorybreakpoint {a}'
'run'

runs to the spot where variable "a" gets modified.

        RAISE exception exceptioninfo

raises that exception in the debugged program.

        ASSIGN varname exp

modifies a variable in the current scope. example:

'assign a a+1'

increases "a" by one.

        WATCH varname varname ...

may be followed by any number of variable names. these will be
inserted into the list of watched variables, which will then
automatically be shown whenever you get to the scope that
contains those variables.

        MEMORY exp

opens a memory window on the address denoted by exp. doesn't
check for validity of the address.

        VARIABLES left top width height

opens the variable view window on that particular spot on the
EDBG screen. If it was already open, it will resize it if
necessary.

        SRCWINDOW srcname left top width height

same for the source view of source "srcname"

        NOABOUT

prevents the about window from popping up

        NOREFRESH

will not refresh memory windows etc. at every step

        REXX n script

sets up the arexx script "script" for launching within EDBG,
from menu item "n".


[note: if you want EDBG to make debugging steps during the
 execution of your scripts, you should launch them with 'rx',
 as those launched from EDBG execute the whole script before
 returning to the debugged program]


@endnode
@node CH_17L
@title "17L. E-VO PreProcessor"
17L. E-VO PreProcessor
--------------------

E-VO has an internal preprocessor which features macro substitution and
conditional compilation. These are not features of the E language,
rather it has been integrated with E-VO for speed and flexibility.

Note that OPT PREPROCESS is not required as the pre-processor is now enabled
by default. It is still supported but does nothing.

Macros.
The macropreprocessor is mostly compatible with Mac2E and the C language PP
(see

You can define macros as follows:

        #define MACRONAME
        #define MACRONAME BODY
        #define MACRONAME(ARG,...) BODY
        #define MACRONAME(ARG,...) BODY \\
          REST OF BODY

MACRONAME and ARG may be ANY case, and may contain "_" and 0-9 as usual.
Whitespace may be added everywhere, except between MACRONAME and "(",
because otherwise E-VO can't see the difference between arguments and a
body.  The BODY may contain occurances of the ARGs.  A macro may continue
on the next line by preceding the end_of_line with a "\\".  A body should
not extend over more than one statement. (a macroname with no body is useful
in combination with conditional compilation).

Macro identifiers have precedence over other identifiers.

Macro definitions defined in a module will be saved in the module only if
OPT EXPORT is on (#define can't be preceded with EXPORT). If that's a
problem, keep macros together in their own modules. Macros in modules can be
used in other code simply by importing them with MODULE and OPT PREPROCESS.


Using a macro.
Using MACRONAME anywhere in the program will insert the body of the macro
at that spot. Note that this substitution is text substitution, and has
little to do with actual E syntax. If the macros have arguments they will
be inserted at their respective places in the macrobody. If the body (or
the arguments) contain futher macros, these will be expanded after that.

example:

        #define MAX(x,y) (IF x>y THEN x ELSE y)

        WriteF('biggest = \\d\\n',MAX(10,a))

will equal writing:

        WriteF('biggest = \\d\\n',(IF 10>a THEN 10 ELSE a))

This immediately shows a danger of macros: since it simply copies
textually, writing MAX(large_computation(),1) will generate code
that executes large_computation() twice. Be aware of this.


Differences of the E PP from the C PP and Mac2E
* it doesn't support 0-arg macros, i.e. MACRONAME()
* comments part of a macro definition are not removed
* macro arguments in string constants are replaced as well. This
  can actually be quite handy, to write macros such as:
  #define debug(x) (WriteF('now evaluating: x\\n') BUT x)

If you try to define the same macro twice you will get an error message.
So there is also a command to undefine a macro if you wish to redefine it.

        #undef MACRONAME

Conditional compilation.
This can be useful if you wish to decide at compile time which part of
your code to use. syntax:

        #ifdef MACRONAME

or

        #ifndef MACRONAME

the piece of source following this will or won't be compiled depending on
whether MACRONAME was defined. you can simply do this with:

        #define MYFLAG

or somesuch. End a conditionally compiled block with:

        #endif

Blocks like these may be nested. example:

#define DEBUG
->#define HEAVYDEBUG

#ifdef DEBUG
  WriteF('now entering bla() with x = \\d\\n',x)
#ifdef HEAVYDEBUG
  WriteF('now dumping memory...\\n')
  /* ... */
#endif
#endif

We can also create more complicated conditional compilation blocks with
#else #elifdef and #elifndef

#ifdef HEAVYDEBUG
  WriteF('now dumping memory...\\n')
  /* ... */
#elifdef DEBUG
  WriteF('now entering bla() with x = \\d\\n',x)
#else
  WriteF('bla()')
#endif

E-VO also defines a default MACRO that you can use to test that you are
compiling with E-VO rather than any other E compiler or even for a specific
version of E-VO.

The currently defined version macro's are

        EVO_3_4_0
        EVO_3_5_0
        EVO_3_6_0
        EVO_3_7_0
        EVO_3_8_0
        EVO_3_9_0
    
So if EVO_3_4_0 is not defined then you are not compiling under E-VO. Further
defines will be added for each new revision of the E-VO compiler so you can
test for a specific minimum version.

Two macros have also been defined to aid with debugging and/or unit testing

  _SRCLINE_
  _SRCPROC_
  
These two macros will be updated during the compilation process to reflect
the current line number and the current PROC name.


        #date - allows you to insert the current date into a a macro.
  
This is useful if you want to define a version string in your code. Acts
like #define but allows the use of date specifiers which are populated
with the compile date/time.

        #date Version '$VER: myApp v0.0 (%d.%aM.%Y) by me'
  
You can then use CHAR to insert that string directly into the executable eg.

CHAR Version,0

The options for the date specifiers are:

        %d - day nr
        %m - month nr
        %y - year nr (4 digits)
        %D - day (name)
        %M - month (name)
        %Y - year nr (2 digits)
        %h - hour (0-23)
        %n - minute (0-59)
        %s - second (0-59)
        %aD - day (abbreviated name)
        %aM - month (abbreviated name)


@endnode
@node CH_17M
@title "17M. E-VO Library mode"
17M. E-VO Library mode
--------------------

NOTE: there are quite a couple of things you need to know before you can
      rush off and create a library. please read this WHOLE chapter before
      proceeding!

E-VO can generate shared library files that can be called from any
programming language, and from any number of tasks at the same time.
Additionally it supports almost all E language constructs in one way or
another.

How?

You can create a library by writing a LIBRARY declaration at the top of the
sourcefile you want to become the library.  syntax:

     LIBRARY name,version,revision,versionstring EXTRA size IS function,...

for example:

        LIBRARY 'mytools.library',37,1,'MyTools 37.1 (2.7.95)' IS
          opengui,closegui,...

The function names are simply those of the PROCs (no methods!) you wish to
make available as functions in the library.  The order in which they appear
is important, because it denotes at which offsets they will be in the
library (starting from -30). You can explicitly assign registers
(e.g. 'myfun(A0)').


A source with a declaration like this is in many respects similar to a
normal E source (i.e.  not a module).  You may also put names of PROCs from
included modules in the LIBRARY declaration.

The EXTRA parameter is optional and reserves an extra amount of space in
the library base struction that you can use for whatever you need. The value
shoudl be 0 - 32000 and also must be even.

If all went ok, E-VO will compile your source to two files:  the library
file (with the name from the LIBRARY declaration, not the source file!), and
a module (".library" replaced by ".m") which contains information for other
E programs on how to use your library

You may now use your library from another E program as follows:

        MODULE '*mymodule'

        PROC main()
          IF mymodulebase:=OpenLibrary('mymodule.library',37)
            /* ... use libfunctions here! ... */
            CloseLibrary(mymodulebase)
          ENDIF
        ENDPROC

or similar.  This is all the E syntax need to create and use libraries,
however, there are quite a number of things one needs to know about what
kind of E code may be put into libraries.


What's so special about libraries then?

Well, imagine that the code in a library is completely separate from the
calling program, and that it may be executed at the same time by two
different tasks.  This weren't so problematic if it weren't for the fact
that code in most languages needs a global environment, not only for storing
global variables, but in E also for quite a lot of features including
exception handling, memory, io, library bases, etc.  (In E, the A4 register
is used for this).  It is safe to say that E code without this specific
global environment is almost impossible.

This environment can't be provided by the program that calls the library,
since this program might be written in C (even for E it would be a problem).
It can't be stored in the library base, since two tasks might be using the
library at the same time.  So what do E libraries do?  They create such an
environment for every task that calls them.  When the library is opened by a
task, a small E environment for that task is created.  When the task makes a
call, the environment belonging to the task is looked up (and put in A4).

In practise this means that libraries created by E should be opened/closed
by any one task EXACTLY ONCE, not more, not less (this shouldn't pose any
problems though).  If you put your library in the PD, you will want to put
a programmers notice in your docs mentioning this.  Strange things WILL
happen if you don't obey this rule.


features and problems/suggestions, hints and tips:
- If you use global variables, or in general the global environment,
  remember that you have a set of global vars for each task that uses your
  library. Global arrays are initialised as normal. (this is what you want,
  in general).
- A procedure main() still needs to be present. It will be called each time
  when a task makes an OpenLibrary() call to you library. Use it only to
  make minor initialisations, i.e. setting global vars to certain values,
  otherwise leave it empty. The code in main() is executed within a
  Forbid()/Permit(). If the source for a LIBRARY contains a PROC close(), it
  will be called when the user of the library calls CloseLibrary() (i.e. it
  is the counterpart to main()). [close still mentioned as unref]
- Lists with variable bits in them, i.e [1,2,a].  A problem when >1 task
  uses the library.  This is a bit of a problem for example in taglists.
  For now one should probably prefix those with a NEW and FastDisposeList()
  them afterwards (hopefully a better solution will be found sometime
  later).
- Exceptions can be raised and caught like normal, only NO exception should
  be thrown out of a library call EVER (strange things might happen). If one
  of your library calls uses code that might throw an exception, you have to
  handle it at the bottom level and turn it into an error code or somesuch.
- In general, and specifically with NEW, if you allocate resources in the
  library, free them in the library. If you feel you have to leave memory
  behind from or to the library, use AllocMem. Also, don't deallocate stuff
  that the calling task allocated for you.
- If you need callbacks, use hooks from the OS. Using pointers to PROCs
  like for example EasyGUI does won't do, as these then will be executed
  within the libraries' environment (i.e. the calling task will not be
  able to reach its own A4 (or equivalent in another language).
- LISP Cells make use of tricky garbage collection techniques on the stack,
  and generally will not work in libraries.  This could be made to work, I
  guess...
- stdin, stdout, conout, arg, wbmessage will all be NIL. If you wish to
  use for example WriteF(), it's best to open (and close!) your own console,
  and place its handle in stdout.
- libraries will stay in memory after use, so make you flush them regularly,
  otherwise debugging becomes particularly hard (you'll be using an old
  version: "didn't I just change that?"). Also close the library in all
  cases, otherwise you'll have to reboot to flush it. A trick to flush 
  directly after testing the library is writing 'AllocMem(100000000,0)' or
  similar directly after the CloseLibrary() call.
- memory allocation like New() and NEW, and other resources that E startup
  code may arrange, are freed when the calling task closes your library,
  i.e. not when your library is removed. Remember that your code might be
  dealing (hopefully transparently) with different sets of global variables,
  NEW memory pools, resources etc.
- If you want to use libraries with other languages like C, note that
  ShowModule has a "-c" option that shows a library module in .fd format,
  which can easily be converted to pragmas for a specific compiler.
- Do not "convert" old code to libraries without looking at it first,
  especially don't put code you don't have the source to in libraries
  without thinking first (it might have static lists, exceptions at weird
  spots etc.). (easygui.m is an example of this).
- Test your code under different circumstances. Have two programs use
  your library at the same time, and close in different orders to see
  wether it has problematic code in it. Try letting it fall over in any
  combination of EDBG, NILCHECK/S, enforcer/mungwall etc. Test your
  code first as normal exe or module before turning it into a library.
- In general, If you do tricky stuff involving the task or global
  environment, remember that you're in a different position from normal
  E code.

Have a look at the examples in the Src/Library dir.

It is currently not possible to create devices directly with E-VO library
link feature. If a lot of people require this feature I will build it in.



@endnode
@node CH_18A
@title "18A. The E grammar"
18A. The E grammar
------------------


This is a grammar of E for those who are interested. Don't expect
it to be up to date or otherwise complete/correct though.

lex syntax: regular expressions
parse syntax: own ASF/SDF adaption;

        name    = grammar ident
        "name"  = constant
        ()      = grouping
        |       = or
        e*      = 0 or more of e
        e+      = 1 or more of e
        {e s}*  = 0 or more of e separated by s
        {e s}+  = 1 or more of e separated by s
        [e]     = e is optional
        ; e     = e is comment :-)



LEX
---



whitespace      = [ \\t]        ; also \\n if last token is [,+-*/] or similar
                  anything between "/*" and "*/"
                  from "->" to \\n
eol             = [;\\n]

constant        = [A-Z] ( [A-Z] [A-Za-z0-9_]* )?
builtin         = [A-Z] [a-z] [A-Za-z0-9_]*
ident,objident  = [a-z] [a-zA-Z0-9_]*

num             = [0-9]+        ; "-" is separate token
                  $[0-9A-Fa-f]+
                  %[01]+
fnum            = [0-9]*.[0-9]*

stringconst     = anything in ''
charconst       = anything in ""



PARSE
-----


program         = opts globalpart localpart

globalpart      = ( modulestat | defstat | objdecl | constdecl | raisedecl )*
localpart       = ( procdecl | constdecl )+

modulestat      = "MODULE" { conststring "," }+ eol
defstat         = "DEF" vardecllist eol
objdecl         = "OBJECT" ident [ "NOALIGN" ] [ "OF" ident ] eol
                     ( vardecllist eol )+
                     [ "UNION" "[" ( vardecllist ) "]" [" ( vardecllist ) "[" "ENDUNION" ]
                  "ENDOBJECT" eol
constdecl       = "CONST" { ( constant "=" constexp ) "," }+ |
                  "ENUM" { ( constant | constant "=" constexp ) "," }+ |
                  "SET" { constant "," }+
procdecl        = [ "EXPORT" ] "PROC" [ "SAFE" ] [ "GETA4" ] [ "STACKCHECK" ] ident "(" argdecllist ")"
                  [ "OF" ident ] ["MODSETUP" | "MODCLEANUP"] ["NOREGS"] [ "HANDLE" ]
                  ( ( "RETURN" | "IS" ) { exp "," }* |
                     eol defstat* stats 
                     [ "EXCEPT" eol stats ]
                     "ENDPROC" { exp "," }* [WITH ("RTR" | "RTE" ]eol )
raisedecl       = "RAISE" { ( constant "IF" builtin "()" compop num ) "," }+
opts            = ( "OPT" { setting "," }+ )*           ; machine dependant

vardecllist     = { vardecl "," }+
vardecl         = ident [ "=" num ]
                    [ ":" ( "LONG" | "PTR" "TO" ptrtype ) ] |
                  ident ":" objtype |
                  ident "[" num "]" ":"
                    ( "ARRAY" |
                      "ARRAY" "OF" ptrtype |
                      "STRING" |
                      "LIST" |
                      ptrtype )
argdecllist     = { argdecl "," }+
argdecl         = ident [ "=" defaultarg ]
                    [ ":" ( "LONG" | "PTR" "TO" ptrtype ) ]
ptrtype         = objtype | simpletype
simpletype      = BYTE | CHAR | INT | WORD | LONG
objtype         = ident

stats           = ( ( onelinestat | multlinestat ) eol )*
onelinestat     = exp |
                  lval ":=" exp |
                  { var "," }+ ":=" exp |
                  lval :=: lval
                  "IF[N]" exp "THEN" onelinestat "ELSE" onelinestat |
                  "FOR" var ":=" exp "TO" exp [ "STEP" num ]
                     "DO" onelinestat |
                  "WHILE[N]" exp "DO" onelinestat |
                  "RETURN" { exp "," }* |
                  "JUMP" ident |
                  "EXIT[N]" exp |
                  "CONT[N]" exp |
                  "WARN" TRUE| FALSE |stringconst |
                  "FATAL" stringconst |
                  "SECTION" ("CODE" | "DATA") ["CHIP"] ["FAST"] | 
                  ( "INC" | "DEC" ) var [ "," num]|               ; nearly obsolete
                  asm_mnemonic { operand "," }* |       ; machine dependant
                  "INCBIN" stringconst |                ; inline asm support
                  simpletype { num "," }+ |
                  "VOID" exp                            ; obsolete
multlinestat    = "IF[N]" exp eol stats
                     [ ( "ELSEIF[N]" exp eol stats )* ] 
                     [ "ELSE" eol stats ]
                     "ENDIF" |
                  "FOR" var ":=" exp "TO" exp [ "STEP" num ] eol
                     stats "ENDPROC" |
                  "WHILE[N]" exp eol stats ["ELSEWHILE[N]" exp eol stats] ["ALWAYS" eol stats] "ENDWHILE" |
                  "REPEAT" eol stats "UNTIL[N]" exp |
                  "SELECT" var eol
                     ( "CASE" exp eol stats )+
                     [ "DEFAULT" eol stats ]
                     "ENDSELECT" |
                  "LOOP" eol stats "ENDLOOP"

explist         = { exp "," }+
exp             = [ "-" ] { [ "NOT" ] item binop }+ |
                  exp "BUT" exp
item            = num | fnum | lval | stringconst | charconst |
                  "SIZEOF" objident |
                  "PSIZEOF" objident |
                  "ARRAYSIZE" [ lval "," ] objident |
                  "OFFSETOF objident.ident |
                  "IF[N]" exp "THEN" exp "ELSE" exp |
                  exp ? exp : exp |
                  exp == [ { (exp | exp "TO" exp) ","}+]
                  "[" explist "]" [ ":" ptrtype ] |
                  ( builtin | ident ) "(" explist ")" |
                  var ":=" exp |
                  "{" ident "}" |
                  "`" exp |
binop           = mathop | compop | logop
mathop          = "+" | "-" | "*" | "/" | "<<" | ">>" | "+=" | "-=" | "*=" | "/=" | "AND=" | "OR=" | "||=" | ">>=" | "<<="
compop          = "=" | "<>" | ">" | "<" | ">=" | "<=" | "=>" | "=<"
logop           = "AND" | "OR" | "ANDALSO" | "ORELSE" | "
constexp        = [ "-" ] { num ( "+" | "-" | "*" | "/" ) }+
lval            = var ( "[" [ exp ] "]" | "." ident )* [ "++" | "--" ] |
                  "^" var [ "++" | "--" ] |
var             = ident
defaultarg      = num



@endnode
@node CH_18B
@title "18B. Tutorial"
18B. Tutorial
-------------


NOTE: the original E v2.1b tutorial has been removed, since it
was not a very useful tutorial, IMHO. Instead, Jason Hulance made
an extensive E tutorial (http://aminet.net/package/dev/e/BeginnersGuideUpd)
that you'd definitely want to take a look at.



@endnode
@node CH_18C
@title "18C. Mapping E to C/C++/Pascal/Ada/Lisp etc."
18C. Mapping E to C/C++/Pascal/Ada/Lisp etc.
--------------------------------------------

[some new stuff has been added here]

In the first/second column I will match E against AnsiC/C++, the third
column is reserved for a third language. I will mainly use Pascal here,
but where a feature asks for it, I will use others (for example, LISP
with quoted expression, Ada with exceptions etc).

note well: take these tables with a grain of salt. I'll try to denote
syntactic equivalences, and semantic properties as well as possible,
but different languages still need their own evaluation.

usage of signs:

-       = feature not available in language in question.
?       = author has no clue what this feature translates to.
          (or atleast he's not sure).
...     = feature may be available, but no appropriate 1:1 translation
          possible to make it interesting.
x,y,z   = arbitrary identifiers
e,f,g   = arbitrary expressions
s,t,u   = arbitrary statements
i,j,k   = arbitrary integers
etc.


-----------------------------------------------------------------------
STRUCTURE/STATEMENTS

E                    C/C++                  Pascal
-------------------- ---------------------- -----------------------
PROC x()             int x() {              FUNCTION x:INTEGER;
PROC x(y,z)          int x(y,z) {           FUNCTION x(y,z:INTEGER):INTEGER;
PROC x(y=1)          int x(y=1) {           -
ENDPROC              return 0; };           x:=0; END;
ENDPROC e            return e; };           x:=e; END;
ENDPROC e,f,g        -                      -
RETURN e             return e;              ?

IF e                 if(e) {                IF e THEN BEGIN
ELSEIF e             } else if(e) {         END ELSE IF e THEN BEGIN
ELSE                 } else {               END ELSE BEGIN
ENDIF                };                     END;
IF e THEN s          if(e) s;               IF e THEN s;
IF e THEN s ELSE t   if(e) s else t;        IF e THEN s ELSE t;

FOR x:=e TO f        - (1)                  FOR x:=e TO f DO BEGIN
FOR x:=e TO f STEP   -                      - (2)
EXIT e               if(e) break;           -
ENDFOR               -                      END;
FOR x:=e TO f DO s   -                      FOR x:=e TO f DO s;

WHILE e              while(e) {             WHILE e DO BEGIN
EXIT e               if(e) break;           -
ENDWHILE             };                     END;
WHILE e DO s         while(e) s;            WHILE e DO s;

s; WHILE e           for(s;e;u) {           s; WHILE e DO BEGIN
  t; u                 t;                     t; u
ENDWHILE             };                     END;

REPEAT               do {                   REPEAT
UNTIL e              } while(!e);           UNTIL e;

LOOP                 for(;;) {              WHILE TRUE DO BEGIN (?)
ENDLOOP              };                     END;

SELECT x             switch(x) {            CASE x OF
SELECT x OF y        switch(x) {            CASE x OF
CASE 1; s...         case 1: s...; break    1: BEGIN s... END
CASE a+1             -                      -
CASE 1,2,3           case 1: case 2: case3: 1,2,3:
CASE "a".."z"        -                      -
ENDSELECT            };                     END

INC x                x++;                   x:=x+1; (INC())
DEC x                x--;                   x:=x-1; (DEC())
JUMP lab             goto lab;              GOTO lab;
x:=e                 x=e;                   x:=e;

/* */                /* */                  { }
->                   //                     -

(1) see WHILE; C has no FOR, "for" in C is another way of writing "while"
(2) only STEP -1 as DOWNTO


-----------------------------------------------------------------------
VALUES

E                       C/C++                   Pascal
----------------------- ----------------------- -----------------------
1                       1                       1
1.0                     1.0                     1.0
$1                      0x1                     ?
%1                      ?                       ?
"a"                     'a'                     chr(97) (?)
'blabla'                "blabla"                'blabla'
[1,2,3]                 - (1)                   -
[1,2,3]:INT             -                       -

(1) in translating from E to C, you can often simulate them with:

myfunc([1,2,3])

becomes:

int dummy [] = {1,2,3};
myfunc(dummy);


-----------------------------------------------------------------------
OPERATORS

E                       C/C++                   Pascal
----------------------- ----------------------- -----------------------
+ - * /                 + - * /                 + - * DIV
= <> > < >= <=          == != > < >= <=         = <> > < >= <=
AND OR  (log)           && ||                   and or
AND OR  (bit)           & |                     ?
SIZEOF x                sizeof(x)               -
`e                      -                       - (1)
^x      (4)             *x                      ...
{x}                     &x                      ...
x++                     x++                     -
x--                     --x                     -
-x                      -x                      -x
IF e THEN f ELSE g      e ? f : g               -
x.y                     x->y                    x^.y
-                       x.y                     x.y
x.y.z                   x->y->z                 x^.y^.z
x:=e                    x=e                     -
e BUT f                 (e,f)                   -
x[]                     x[0] *x (2)             x[0]
x[1]                    x[1]                    x[1]
x[1]    (3)             &x[1]                   ?
x[1].y                  x[1]->y                 x[1]^.y
x[]++                   *x++                    -
x[1].y++                *(x+1)++                -
x::y.a                  ((y *)x)->a             -
x.y::z.a                ((z *)x->y)->a          -

(1) see QUOTED EXPRESSIONS
(2) also for others, equivalences between *(x+e) and x[e] hold.
(3) if ARRAY OF <object>
(4) ONLY for giving by reference. otherwise: "[]"

-----------------------------------------------------------------------
CONSTANTS/TYPES

E                       C/C++                   Pascal
----------------------- ----------------------- -----------------------
CONST X=1               #define X 1             CONST X=1;
                        const int X=1;
ENUM X,Y,Z              #define X 0 (etc.)      TYPE x=(X,Y,Z);
                        enum x{X,Y,Z};
SET X,Y,Z               -                       TYPE x=SET OF (X,Y,Z);

DEF                                             VAR
x                       int x; (or: long x;)    x:INTEGER;
x:LONG                  int x;                  x:INTEGER;
x:PTR TO y              struct y* x;            x:^y;
x:y                     struct y x;             x:y;
x[10]:ARRAY OF y        struct y x[10];         x:ARRAY [0..9] OF y;
x[10]:STRING            - (1)                   x:STRING[10]; (2)
x[10]:LIST              - (1)                   - (1)

x:REG                   register int x;

OBJECT x                struct x {      (3)     TYPE x = RECORD
  y:CHAR,z:INT            char y; short z;        y:CHAR; z:INTEGER;
ENDOBJECT               };                      END;

(1) when translating from E to C, simulate with an array of char/int resp.,
    and do your own range-checking etc.
(2) no Wirth Pascal, but available in all popular dialects.
(3) or public class.


-----------------------------------------------------------------------
QUOTED EXPRESSIONS

E                       LISP                        MIRANDA
----------------------- --------------------------- -------------------
`e                      (QUOTE e)  'e               (3)
                        (LAMBDA () e)     (1)
`x+y                    '(+ x y)
Eval(`e)                (EVAL `e)
ForAll(v,l,`e)          - (2)
MapList(v,l,l,`e)       (MAPCAR (LAMBDA (V) E) L)   map (\\v->e) l

example:

E:              MapList({x},[1,2,3,4],a,`x*x)
MIRANDA:        map (\\x->x*x) [1,2,3,4]
LISP:           (MAPCAR (LAMBDA (X) (* X X) `(1 2 3 4))

(1) really QUOTE, but sometimes used where in LISP LAMBDA would be
    used, like in MapList()
(2) not even in ProLog, see other logical languages.
(3) lazyness would be used instead here

-----------------------------------------------------------------------
UNIFICATION AND LISP CELLS

E                       LISP                    PROLOG
----------------------- ----------------------- -----------------------
<1|2>                   (1 . 2)                 [1|2]
<1,2,3>                 (1 2 3)                 [1,2,3]
<1,2|3>                 (1 2 . 3)               [1,2|3]

E                       HASKELL                 PROLOG
----------------------- ----------------------- -----------------------
e <=> <x|y>             (x:y) = e               e = [X|Y]
e <=> <1,2,x>           [1,2,x] = e             e = [1,2,X]
e <=> [1,x]             -                       -


-----------------------------------------------------------------------
EXCEPTIONS

E                       C++                     ADA
----------------------- ----------------------- -----------------------
PROC x() HANDLE         int x() { try {         function x is begin
EXCEPT                  } catch (exc) {   (1)   exception
EXCEPT DO               -                       -
ENDPROC                 }};                     end x;

Raise(e)                throw e;                raise e;
Throw(e,f)              ?                       -
ReThrow()               throw e;                raise e;
RAISE "MEM" IF New()=0  -                       - (2)


(1) catch handles only one specific exception, it's quite different
    from general exception handlers as used in E.
(2) the runtime system does raise some exceptions, but I'm not sure
    whether automatically raised exceptions can be _defined_ in Ada.


-----------------------------------------------------------------------
OBJECT ORIENTED PROGRAMMING

E                               C++
------------------------------- -----------------------
OBJECT x                        class x {
OBJECT x OF y                   class x : y {
self.i                          this->i
PROC a OF x IS self.i           virtual int x::a() { return i; }
-                               int x::a() { return i; }
PROC a OF x IS EMPTY            virtual int x::a() =0
PUBLIC                          public:
a.method(1)                     a->method(1)
SUPER a.method(1)               a->supername::method(1)

[see also next part under NEW]

-----------------------------------------------------------------------
BUILTIN FUNCTIONS AND MEMORY ALLOCATION
(only a few are presented here, as an example)

E                       C/C++                   Pascal
----------------------- ----------------------- -----------------------
WriteF(fs,...)          printf(fs,...);         WriteLn(a,b,...);
                        cout << a << b ... ;

ReadStr(f,s)            scanf(fs,...)           ReadLn(s)
Val(s)                                          Val()
                        cin >> s;

StrCopy(s,s,n)  (1)     strcpy(s,s)             s:=s; (2)

Mod(e,e)                e%e                     e MOD e
Shl(e,n)                e<<n                    Shl()
Long(e)                 -                       -


p:=New(e)               p=malloc(e);            New(p);
NEW p                   p=new type;
NEW p.constr()          p=new constr()          -
NEW [e,f,g]             -                       -
Dispose(p)              free(p);                Dispose(p);
END p                   delete p;


(1) when translating from C, make sure you turn the arrays of char into
    proper STRINGs.
(2) dunno what function is needed in the pointer case.


@endnode
@node CH_18D
@title "18D. Amiga E FAQ"
18D. Amiga E FAQ
----------------
[94_10_1..94_12_1]

This FAQ-list (Frequently Asked Questions) was compiled by looking through
old email and gathering those questions which keep popping up.



Compiler/Linking/Executables

- Will E-VO generate code for PowerPC Amiga's, if they become available?

  Likely, though don't pin me down on this. I have an idea on how to
  adapt E-VO for PPC, and maybe other processors/languages as well.
  When, where and how this is going to happen depends, and you'll just
  have to wait. A project like this easily requires a year's worth
  of hard work, so if I'm to go through with it, it'd better be worth it.
  That of course depends heavily on the future of the Amiga.

- How can I link E code with other languages? / use standard object (.o)
  format?

  Converting between .m and .o really isn't a huge problem, as one can see
  from the o2m utility, which allows cooperation between E and
  Macro-Assembly quite marvelously. The problem with C is in the compiled
  code, not the file format. Anything but trivial C will reference things
  like _DOSBase, link- library functions, stack-bases or other things from
  startup-code that E provides in a quite different way, for example E's
  startup code is very different from that of C compilers, and dosbase in E
  is placed on the stack,not in the exe as with C. Even two C compilers may
  have these problems. If you can get your C compiler to deliver a .o
  without external references etc., you can link it with E (I did this once
  with MaxonC++). If you need to program a project in both E and another
  language, you may wish to look at E's (or the other language's) library
  linking facilities. E is really quite a bit different from other languages
  (e.g. C), and providing .o linking facilities only makes sense if the code
  contained in the .o files can cooperate, which isn't the case with E.

- Can I link amiga.lib?

  Currently no. To a lesser extend this has the same problems as .o files.
  Many functions from amiga.lib have already been "converted" to E modules.
  (see

- Can E code be made resident?

  Nearly ALL E code is already resident-able without the help of the
  programmer. As far as I know, the only E construct that can violate this 
  is a static list [] with an expression in it (so [a], for example. [1] or
  NEW [a] is ok.). There are also a few built in functions that are not pure.
  Use OPT PURE if you want to ensure these are not present in your code.

- I have this application with pieces of inline asm, and after
  OPTI/S it behaves weird. what is going on?

  Inline asm may use the same registers as the register optimizer does.
  check E.doc for this.

- I wrote X in both C and E, and the E version is Y times slower/faster.
  how come?

  It's not easy to compare both in a fair way. Often it may be the case that
  one of the two has more optimized routines for something (string handling,
  I/O etc.). Use of a profiler can reveal this. If the code in question only
  does calculations, there's no reason why any of the two should be
  significantly slower than the other, if used properly.

- I have written this nice 'myutils.m', only when I use one function
  from it, E-VO links all. I don't like that.

  The style of writing modules in E is to keep them as relatively small
  units, with only related code and data. (This makes more sense if you know
  that the module is the unit of datahiding in E). Split it up!

- Can I make E-VO resident?

  no, currently not. E-VO itself is still written in old-fashioned assembly
  style, and cannot be made resident ;-)

- Can we have an option for E-VO to generate assembly source instead of an
  executable?

  No. There's no intermediate level of assembly in E-VO, it generates code
  directly. Using a disassembler is as close as you will get.

- I have been disassembling some code generated by E-VO, and I think this
  bit of code here can be optimized to this bit of code.

  Sure it can, I can see that too. But E-VO is not an assembly programmer.
  Doing really very high quality optimisations would require a complete
  rewrite of E-VO, and would make it a lot slower, even with optimisation
  off. Writing a good optimizer for a compiler is quite a non-trivial
  operation, and if some E-VO generated code is sub-optimal or downright
  silly doesn't mean I must be amateur assembly programmer too, and you need
  to explain me how its done.

- E should have a CHIP keyword or somesuch, to put data areas in chip
  memory.
  
  E-VO does now have a SECTION command which allows this.

Resources

- There's no explanation in the docs of all those handy functions from
  intuition.library etc. How come?

  The functions are not part of E, rather they are part of the amiga OS.
  As such, they are described by Commodore's documents, not by the E
  documents. "The AMIGA ROM Kernal Reference Manuals" are a series of books
  published by Addison Wesley. A good place to start is for example
  "Libraries", ISBN 0-201-56774-1. Other books are worth considering too,
  an example is "The Amiga Guru Book" by Ralph Babel. All the C code you
  come across in the RKRMs is available in E form in Src/RKRM/

- I'd like to read more source code than what comes with the distribution
  where should I look?

  There are lots of places too look, but the best is without doubt Aminet
  (A collection of FTP-sites on the internet), for example ftp.luth.se.
  In the directory 'pub/aminet/dev/e' you'll find all sorts of E related
  stuff. Some BBS'es outside of the internet also carry Aminet, and there
  are even CDROMs available. Another good place to pick up sources and
  to talk with fellow E programmers is the E mailing list. Send an email
  with 'SUBSCRIBE' in the subject to 'amigae-list@intercom.it' (or
  'UNSUBSCRIBE' if you want get off it again), and any future messages to
  the same address. If you have any problems with the list email
  'fsoft@intercom.it' (Fabio Rotondo, the list administrator). The old list
  at amigae-request@bkhouse.cts.com is no longer operational.
  There are E discussion on other nets as well (FidoNet, AmigaNet),
  public domain disk series (EPD in europe/germany, same address as the
  german  registration site), furthermore there are heaps of user-clubs,
  BBS's etc. supporting E these days. just look around in your area.

- I have troubles understanding your incomprehensible english. Can I read
  about E in my own language?

  There are translations of E.guide (some older versions) in german, french,
  spanish and italian, the first three are available on Aminet.

- Where can I find any E related sites on the WWW?

  The wikipedia page is probably the best starting point. It has some useful
  links: 
  
  https://en.wikipedia.org/wiki/Amiga_E

- I have heard about this mailing list. is it any good?

  Find out for yourself. If you're serious about E it's definitly a must,
  also because it's the place where news on E is generally released first.
  Many good E programmers are on it as well, so you are sure to get the
  right answers to your questions.

- I have a hard time getting on the mailing list. Could you take care of
  that for me?

  I personally don't have anything to do with the administration of the
  list, so I can't help out any better than anyone else can. Remember that
  the server is an automated process, so you need to be very precise in what
  you send there. Don't send administrative post to the list at large,
  instead try and get hold of the administrator
  (Fabio Rotondo: fsoft@intercom.it).

- I'm new to internet. Could you help me get on the mailing list /
  FTP files from Aminet etc...?

  Stuff like that is way beyond my service. Please try and contact someone
  locally.

- How do I know what is the latest version / How do I receive it?

  If you're on the mailing list you'll be the first to know. Aminet is the
  place where releases and updates are uploaded first.

- Can you send me email as soon as a new version is out?

  No. I used to send all registered users email with new versions, but can't
  do so anymore as the number is too large (last time I tried I jammed our
  network and crashed various workstations).

- How do I register?

  No need to register.

- Does E exist for other platforms?

  Not Yet. Personally I have undertaken small projects to make portable
  compilers/translators, and several others have E compiler projects on
  other platforms, but nothing has come out of this so far. Portability
  of some form might be available as a spin-off of the PPC backend project

- Can I become an E registration site for country X?

  No longer applicable.

- Can I start an E support BBS / an E programming club...?

  You can always do that, without asking me. I of course enjoy hearing about
  efforts like this...

- Just out of curiosity, how many registered E programmers are there?

  This changes all the time. Inquire within :-)


Programming

- Can I do X in E? (where X = {games,dtp-packages,...})

  E is a general purpose programming language, so there shouldn't be any
  type of program that can't be done in E (with few rare exceptions, which
  are covered by E's inline asm). This doesn't mean E is especially equipped
  for certain types of programs, i.e. E has no special functions for games
  (though its extensibility makes it easy to add them).

- How do I create X in E (where X = {window, interupt handler,...})

  as with the last question, E just opens the possibilities to write
  anything, and there isn't always a specific function available. In the
  worst case this means having to use difficult functions from spooky
  libraries, but it's worth it getting to grips with that. If you're lucky
  some E programmer already has done something similar which you can learn
  from.

- Please write me an example how to do X in E.

  Please try to figure it out yourself first, or ask other E programmers to
  help out.

- The compiler refuses to compile things like myscreen.rastport.bitmap,
  eventhough x.y.z is generally possible. Why is that?

  This should not be a problem anymore. The only case where this can still
  occur is if you do not include the module that defines the object "z" is
  in.

- I'm writing code like this:

        DEF s[100]:STRING
        s:='my beautiful string'

  which is allowed by the compiler, but later gives me problems. What can
  possibly be wrong with this?

  If you're used to for example BASIC, you're used to handle strings in the
  same way as integers, as they are both _values_. In E however there are no
  real string variables in the BASIC sense, the DEF above creates a piece of
  memory to hold a string, then sets the variable as a pointer to this
  memory. The pointer and the memory are implementation-wise two unrelated 
  entities.From that DEF on, all operation that directly access 's' access
  the pointer, which is just an integer which tells us were to find the
  actual string. The assignment thus puts the address of 
  'my beautiful string' into 's', overwriting the old value. The 
  string-memory created by the DEF sits there unaltered, and now 
  unaccessable because no pointer points to it. Functions like StrCopy()
  can use this pointer to find the real memory, and can fill it:

        StrCopy(s,'my beautiful string')

  is correct. Of course assigning strings as pointers also has a use, for
  example if you just want to read the string data and do nothing else with
  it. then:

        DEF s:PTR TO CHAR

  allocates no memory for a string, only the pointer. the assignment above
  would now make sense. This all boils down to the differences between
  pointers and values (or value/reference semantics), and is important to
  understand to succesfully program in E.

- How do I return a STRING / OBJECT etc. from a function?

  depends. If you just wish to use it as temporary return value, give a
  string as argument and let the routine fill it. If you need to create a
  NEW string/ object, allocate it dynamically within the function, and
  return the pointer.

- can I create an array of typed pointers, eg ARRAY OF PTR TO LONG

  Yes, it was added in E-Vo 3.7.0 and futher improved in 3.8.0

<more of these specific questions are to be added>



Bugs

- I remember programming something when suddenly the compiler crashed/
  misbehaved/produced an internal error/gave wrong error messages. shouldn't
  you be doing something about this?

  If I have no clue where to look for a bug I can't fix it. If you stumble
  across something that you're sure of is a compiler bug, make a copy of
  the source that reproduces the bug (possibly cut the source down while
  making sure it still shows the bug) and send it to me with as much info
  as possible. enforcer hits for example are very useful.

- I wrote program X, but it crashes. I'm positive I made no mistakes,
  so certainly this must be a compiler bug.

  due to E's untyped-ness, you can never be quite sure you're code is
  correct. Most code send to me with comments like the above later proved
  to be errors in the code, mostly due to lack of E knowledge. Read the
  docs, and use EDBG!



Future Features

Will E have...
- Bitfields?
- Startup code that doesn't open intuition etc.?
- A feature to automatically include all modules used by a module?
- Access to object members in methods without the 'self.' prefix?
- DEFs halfway PROCs or local to loops etc.?
- Disassembly mixed mode for EDBG?
- Sourcecode level #include?
- Unsigned LONG?
- Multiple statements in macros?
- Multiple error messages?
- Dereferenced variables in lefthandsides of assignment expressions,
- multiple returnvalue assignments, and the {} operator?

No. Or atleast very unlikely.

- Registerised arguments?

  It surely is possible but like many things it's not a priority for me.

- Library linker?

  Yes, it's included in this version.

- Multiple Inheritance?

  Not possible in E. MI breaks with structural compatability between
  objects, and needs relatively strong typing to resolve this.

- PROCs inside PROCs?

  Not likely.

- multidimensional arrays?

  Included in E-VO 3.5.0

- Overloading

  No. Overloading requires quite strong typing to distinguish which function
  to call. Besides, overloading is 100% syntactic sugaring, it doesn't
  provide any added functionality whatsoever.

- more 020/881 support?

  Yes, eventually. It just isn't at the top of my TODO-list.

- assembly source output?

  No. Unlike other compilers, E never passes an assembly stage, E code is
  compiled directly to opcodes. Providing this feature would boil down
  to doing a disassembly. E's inline assembler is not used for generating
  code for E constructs. (see

- C compatible syntax?

  Often people are used to some type of language (mostly C), and don't
  understand why a programming language design isn't as configurable as a
  directory utility or text editor:
  'I like "=="/"=" better than "="/":="'
  'Can't you switch """ and "'" ?'
  'Why doesn't "--" work as in C?'
  'I hate typing CAPS for keywords!'
  'No precedence sucks!'
  The features 'referenced' above will never change, and you'd better get
  use to them. All design decisions have good reasons, and unlike some
  rumours they were not done for compiler speed (infact implementing them
  more traditionally wouldn't make the compiler any slower).

- Why is feature X in E not like language Y, which I find better.

  Like above, the design of a language isn't changed just because of
  personal preference. If you have strong ideas of what a language should
  look like according to you, and those ideas stem from language Y, then
  use Y. If you can't find your ideas in any existing language, then it's
  time to design and implement your own! (I'm not kidding, it's worth the
  trouble! and you can always start with E... :-)

- X maybe?

  Having seen literally hundreds of programming languages, your chances of
  suggesting a new feature for E which I haven't thought of before are not
  too bright. Many things don't 'fit' into E, and in general time is limited
  to add more fancy stuff. (If you look closely, the amount of features in
  E is already quite high for an average programming language). I have quite
  a huge list of possible features for E, and eventually some of these will
  be implemented. just wait and see...




@endnode
@node CH_18E
@title "18E. TODO/BUG list"
18E. TODO/BUG list
------------------
By popular demand:  This is an extract from my todo/bugs list for E.  The
real list is much longer, since I left all the real details/unimportant
stuff out. Don't go and ask why any of these items haven't been done yet;
the answer will generally vary from the the tactical "due to the internal
structure of the compiler fixing this bug is more work than the collective
trouble it will ever cause" to the honest "I'm too lazy: I can't be
bothered" :) Don't go and ask me "when?" either.

        Bugs I haven't been able to reproduce:
*       compilation problems on 060
*       broken Lock in combination with MCX

        Bugs still not fixed:
*       dereferencing an untyped OBJECT member is not properly interpreted
        as "PTR TO CHAR". Easy workaround: Give the member a type.
*       error reports can sometimes be off by several lines.
*       error reports for inline assembly often doesn't work.
*       unreferenced check too conservative in some cases.
*       doesn't detect two library functions with the same names.
*       constant folding optimizer often misbehaves, disallowing
        you to do things like '1+SIZEOF o' in constant expressions
        or 1+a as list-elements.
*       internal linker problem doesn't allow for private methods.
*       strings in lists cannot be joined by "+" (optimizer problem).
*       StrCopy/MidStr don't fill dest with empty-string if len=0
*       methods and procs in one module cannot have the same name
        (other than by inheritance).
*       Exists() evaluates all regardless.
*       SUPER doesn't check wether the superclass actually has such
        a method.
*       recompiling a module that is being used at the same time
        by another invocation of E-VO can give problems.

        Future Todo's
*       fix above bugs :)
*       extend CONST expressions with floats etc.
*       find a way to split up executables in more hunks
*       add symbols for methods
*       lots of extensions to the optimizer
*       proper 020+ support
*       improve error reporting
*       class variables

        Wild Ideas (aka Big Todo's)
*       plug a general back-end onto E-VO (to allow for code
        generation to other cpus (e.g. PPC) or even languages
        (e.g. C) or better optimizers etc.)
*       extend the language with lambda's / higher order
        functions / lazyness etc.
*       concurrency system based on Linda
*       better debugging / analysis / type inference tools
*       full visual language on top of E



-------------------------------EOF--------------------------------------
@endnode