Counting bytes in binary variable

[dnewhall]

Is there any way to get the number of bytes in a binary variable without using wc?

typeset -b bin
read -n 10 bin < /dev/zero
print ${#bin}
16
print $bin
AAAAAAAAAAAAAA==
print $(printf %B bin | wc -c)
10

[hyenias]

Sure. Here is a solution using ksh only and thus avoiding external programs like wc.

ksh93 displayes the results of a binary varible in base64. Base64 encoding of a binary varabile will consume four base64 encoding characters for every 3 bytes of the binary variable. Padding of the unused bits is represented by an equal sign.

All that is needed to determine the count of bytes that were encoded is subtract the number of padding chars from the base64 encoded string length, divide by 4, multiple by 3, then floor the result. [ 1/4 * 3 = 3/4 = .75 ]

Continuing with your example setup:

Version 1: Base64 binary variable byte count

$ unset bin; typeset -b bin
$ read -n 10 bin < /dev/zero
$ typeset -p bin
typeset -b bin=AAAAAAAAAAAAAA==''
$
$ true ${bin%%=*}; echo $(( floor( (${#bin} - ${#.sh.match}) * .75) ))
10

Version 2: Base64 binary variable byte count storing result in short integer x avoiding the floor as conversion to integer truncrates the fraction.

$ unset b x; typeset -si x; typeset -b b; read -n 22 b <<< 'Kornshell is the Best!'
$ echo $b | base64 --decode; echo                                                    
Kornshell is the Best!
$ true ${b%%=*}; (( x = ( ${#b} - ${#.sh.match} ) * .75 )); typeset -p b x           
typeset -b b=S29ybnNoZWxsIGlzIHRoZSBCZXN0IQ==''
typeset -s -i x=22
$ 

What stuff does:

• true ${b%%=*}: performs a variable expansion to obtain a matching result of all the padding characters
• ${#.sh.match}: count the length of the just recently issued variable expansion containing padding characters if any

[dnewhall]

Oh, duh. I should have thought about just reading the base64 description and mathing it out. I'm now using a variant of your first version in some code.

However, I did find there is still one case I've found where piping to wc is still the best option: if you don't know whether a variable holds string or binary data.

function byte_count {
# byte_count varname
    nameref var=$1
    print $(printf "%B" var | wc -c)
}

typeset txt="This is some text!"
typeset -b bin
read -n 1024 bin << "This is some text as binary!"
byte_count txt
18
print ${#txt}
18
byte_count bin
29
print ${#bin}
40