Version 3.2.0 - October, 2024

• Discontinued support for Python 3.6, 3.7, and 3.8. Adopted new Python features from Python versions 3.7-3.9:

  • Updated type annotations to use built-in container types instead of names imported from the typing module (e.g., list[str] vs List[str]).
  • Reworked portions of the packrat cache to leverage insertion-preserving ordering in dicts (including removal of uses of OrderedDict).
  • Changed pdb.set_trace() call in ParserElement.set_break() to breakpoint().
  • Converted typing.NamedTuple to dataclasses.dataclass in railroad diagramming code.
  • Added from __future__ import annotations to clean up some type annotations. (with assistance from ISyncWithFoo, issue Changes towards 3.2.0 #535, thanks for the help!)

• POSSIBLE BREAKING CHANGES

  The following bugfixes may result in subtle changes in the results returned or exceptions raised by pyparsing.

  • Fixed code in ParseElementEnhance subclasses that replaced detailed exception messages raised in contained expressions with a less-specific and less-informative generic exception message and location.

    If your code has conditional logic based on the message content in raised ParseExceptions, this bugfix may require changes in your code.

  • Fixed bug in transform_string() where whitespace in the input string was not properly preserved in the output string.

    If your code uses transform_string, this bugfix may require changes in your code.

  • Fixed bug where an IndexError raised in a parse action was incorrectly handled as an IndexError raised as part of the ParserElement parsing methods, and reraised as a ParseException. Now an IndexError that raises inside a parse action will properly propagate out as an IndexError. (Issue IndexErrors are handled differently than other exceptions #573, reported by August Karlstedt, thanks!)

    If your code raises IndexErrors in parse actions, this bugfix may require changes in your code.

• FIXES AND NEW FEATURES

  • Added type annotations to remainder of pyparsing package, and added mypy run to tox.ini, so that type annotations are now run as part of pyparsing's CI. Addresses Issue Type checking as part of the CI pipeline #373, raised by Iwan Aucamp, thanks!

  • Exception message format can now be customized, by overriding ParseBaseException.format_message:

    def custom_exception_message(exc) -> str:
        found_phrase = f", found {exc.found}" if exc.found else ""
        return f"{exc.lineno}:{exc.column} {exc.msg}{found_phrase}"
    
    ParseBaseException.formatted_message = custom_exception_message
    

    (PR Add found string as property #571 submitted by Odysseyas Krystalakos, nice work!)

  • run_tests now detects if an exception is raised in a parse action, and will report it with an enhanced error message, with the exception type, string, and parse action name.

  • QuotedString now handles translation of escaped integer, hex, octal, and Unicode sequences to their corresponding characters.

  • Fixed the displayed output of Regex terms to deduplicate repeated backslashes, for easier reading in debugging, printing, and railroad diagrams.

  • Fixed (or at least reduced) elusive bug when generating railroad diagrams, where some diagram elements were just empty blocks. Fix submitted by RoDuth, thanks a ton!

  • Fixed railroad diagrams that get generated with a parser containing a Regex element defined using a verbose pattern - the pattern gets flattened and comments removed before creating the corresponding diagram element.

  • Defined a more performant regular expression used internally by common_html_entity.

  • Regex instances can now be created using a callable that takes no arguments and just returns a string or a compiled regular expression, so that creating complex regular expression patterns can be deferred until they are actually used for the first time in the parser.

  • Added optional flatten Boolean argument to ParseResults.as_list(), to return the parsed values in a flattened list.

  • Added indent and base_1 arguments to pyparsing.testing.with_line_numbers. When using with_line_numbers inside a parse action, set base_1=False, since the reported loc value is 0-based. indent can be a leading string (typically of spaces or tabs) to indent the numbered string passed to with_line_numbers. Added while working on set_parse_action handler receives seemingly erroneous loc argument #557, reported by Bernd Wechner.

• NEW/ENHANCED EXAMPLES

  • Added query syntax to mongodb_query_expression.py with:

    • better support for array fields ("contains", "contains all", "contains any", and "contains none")
    • "like" and "not like" operators to support SQL "%" wildcard matching and "=~" operator to support regex matching
    • text search using "search for"
    • dates and datetimes as query values
    • a[0] style array referencing

  • Added lox_parser.py example, a parser for the Lox language used as a tutorial in Robert Nystrom's "Crafting Interpreters" (http://craftinginterpreters.com/). With helpful corrections from RoDuth.

  • Added complex_chemical_formulas.py example, to add parsing capability for formulas such as "3(C₆H₅OH)₂".

  • Updated tag_emitter.py to use new Tag class, introduced in pyparsing 3.1.3.