Pyparsing is a pure-Python class library for quickly and easily constructing recursive-descent parsers. Pyparsing takes a "building-block" approach to parser construction, as opposed to code generation methods (such as lex/yacc) or pattern definition strings (such as regular expressions).
Version 1.3.1 includes some minor enhancements, plus some performance improvements that can really improve performance for grammars that use the Combine class (often used in specifying floating point numbers). The 1.3.1 change notes are listed below. Download pyparsing at http://pyparsing.sourceforge.net. -- Paul McGuire Version 1.3.1 - June 12, 2005 ----------------------------- - Added markInputline() method to ParseException, to display the input text line location of the parsing exception. (Thanks, Stefan Behnel!) - Added setDefaultKeywordChars(), so that Keyword definitions using a custom keyword character set do not all need to add the keywordChars constructor argument (similar to setDefaultWhitespaceChars()). (suggested by rzhanka on the SourceForge pyparsing forum.) - Simplified passing debug actions to setDebugAction(). You can now pass 'None' for a debug action if you want to take the default debug behavior. To suppress a particular debug action, you can pass the pyparsing method nullDebugAction. - Refactored parse exception classes, moved all behavior to ParseBaseException, and the former ParseException is now a subclass of ParseBaseException. Added a second subclass, ParseFatalException, as a subclass of ParseBaseException. User-defined parse actions can raise ParseFatalException if a data inconsistency is detected (such as a begin-tag/end-tag mismatch), and this will stop all parsing immediately. (Inspired by e-mail thread with Michele Petrazzo - thanks, Michelle!) - Added helper methods makeXMLTags and makeHTMLTags, that simplify the definition of XML or HTML tag parse expressions for a given tagname. Both functions return a pair of parse expressions, one for the opening tag (that is, '<tagname>') and one for the closing tag ('</tagname>'). The opening tagame also recognizes any attribute definitions that have been included in the opening tag, as well as an empty tag (one with a trailing '/', as in '<BODY/>' which is equivalent to '<BODY></BODY>'). makeXMLTags uses stricter XML syntax for attributes, requiring that they be enclosed in double quote characters - makeHTMLTags is more lenient, and accepts single-quoted strings or any contiguous string of characters up to the next whitespace character or '>' character. Attributes can be retrieved as dictionary or attribute values of the returned results from the opening tag. - Added example SimpleCalc.py, a refinement on fourFn.py that adds an interactive session and support for variables. (Thanks, Steven Siew!) - Added performance improvement, up to 20% reduction! (Found while working with Wolfgang Borgert on performance tuning of his TTCN3 parser.) - And another performance improvement, up to 25%, when using scanString! (Found while working with Henrik Westlund on his C header file scanner.) - Updated UML diagrams to reflect latest class/method changes. -- http://mail.python.org/mailman/listinfo/python-list
