Release 0.21.0

Changed

• Transient state is cleared after parse to lower memory usage.
• Remove direct parser references from LRStackNode and GSSNode context objects,
  making them more self-contained and easier to reason about. See e51e5ffa.
• Update error_recovery signature to receive default_error_recovery as a
  parameter instead of accessing it via context.parser. See e51e5ffa.
• Store symbol class reference in symbol.cls for faster access. See e51e5ffa.

Full Changelog: 0.20.0...0.21.0

Release 0.20.0

See CHANGELOG.md

0.19.0

Full Changelog: 0.18.0...0.19.0

Release 0.18.0

See the changelog

Release 0.17.0

See the changelog

Release 0.16.1

Fixed

• Fix AST to_str in the context of syntactic sugar (BNF extensions) 6a99f96.

0.16.0

Added

• _pg_extras slot on dynamically created Python objects for additional
  user-defined information (e.g. info used during semantic analysis) (#140)
• _pg_children/_pg_children_names on dynamically created Python objects for
  named assignments 0771ca5, 204b5a0.
• to_str method for dynamically created Python objects 204b5a0.

Changed

• visit function of the visitor pattern now has a third parameter which is the
  depth of the current tree node [63e6e42]. (BIC)
• Relaxed click dependency to allow for 8.x versions. (#142). Thanks GabDug@GitHub.

Fixed

• Fix visitor memoization caching #138. Thanks vladaindjic@GitHub.

0.15.0

Added

• forest.get_first_tree() for efficiently unpacking tree 0 7cb006c.
• forest.disambiguate(callable) for removing ambiguities from the forest f4664af.
• Greedy repetitions (*!, +!, and ?!) 517d33e.

Changed

• Make Parent iterable over possibilities 57606c9.

Fixed

• Fix parse tree iteration and indexing 6947942.
• Fix passing sub-results in dynamic filter 65a1cbf.

0.14.0

This release brings multiple new features and improvements. Read the Release Notes for more info.

Added

• pglr parse for parsing files and displaying parse trees and forests.
• pglr trace --frontier for organizing GSS nodes for GLR trace into
  frontiers (a.k.a. shift levels)
• to_dot on trees and forests for rendering GraphViz dot string. Used in
  pglr parse if --dot switch is provided to create dot file of a parse
  forest/tree.
• Parenthesized groups in grammar rules.
• New examples: JSON, BibTeX, Java (based on Jave SE 16).
• New performance tests based on the new example grammars.

Changed

• Changed tree_str on parse trees to to_str. (BIC)
• Improved GLR implementation. Improved performance.
• GLR parsing now returns Forest object which can be indexed and iterated
  yielding lazy trees. See Release Notes for more info. (BIC)
• Dropping support for deprecated Python 3.4 and 3.5 versions. (BIC)

0.13.0

Changed

• Optimized calculating line/column from a position (#131). Thanks xor2003@GitHub.
• LR/GLR implementation rework and cleanup. Corrected handling of EMPTY
  reductions. Support for full CFG set. Better tracing output. (possible (BIC))
• Removed context parameter to parse* calls. Added extra parameter for
  keeping of additional state during parsing. If extra is not given it is dict by default. ((BIC))
• Dynamic disambiguation filter parameters change ((BIC), see the docs)
• Removed explicit Context class. Now, there are context-like objects (
  parser head, error context, GSS node parent etc.) (possible (BIC))
• Moved project meta-data to setup.cfg and introduced git based versioning
  using setuptools_scm. Thanks KOLANICH@GitHub (#104).

Fixes

• During GLR rework several bugs are fixed:
  • Wrong positions reported with GLR and EMPTY. Reported by Hyldrean@GitHub
    (see #110)
  • Issue with handling of EMPTY productions in ambiguous grammars. Reported
    by johnw3d@GitHub and stuartlangridge@GitHub (see #112, #114)