o extract_python fixes: - now returns None for non-string arguments - no longer extracts strings from nested function calls refs #38 - use the correct starting line number in multi-line gettext function calls - avoids falsely identifying string keyword arg defaults from function definition names that match a keyword, e.g.: def gettext(foo='bar') - avoid capturing translator comments embedded within a gettext function call - default the file encoding to iso-8859-1 instead of ascii when missing a magic encoding comment, to emulate pre Python 2.5 behavior. Python warns about 'non-ascii' chars when there is no magic encoding comment, but < 2.5 actually treats them as iso-8859-1 for backwards compat (PEP 263). >= 2.5 treats them as strict ascii o extract fixes: - filter out messages that don't contain strings where the keyword specification calls for fixes #39 - filter out empty string messages and emit a warning about them, like xgettext

--- a/ChangeLog
+++ b/ChangeLog
@@ -12,6 +12,13 @@
  * The number formatting functions now also work with numbers represented by
    Python `Decimal` objects (ticket #53).
  * Added extensible infrastructure for validating translation catalogs.
+ * Fixed the extractor not filtering out messages that didn't validate against
+   the keyword's specification (ticket #39).
+ * Fixed the extractor raising an exception when encountering an empty string
+   msgid. It now emits a warning to stderr.
+ * Numerous Python message extractor fixes: it now handles nested function
+   calls within a gettext function call correctly, uses the correct line number
+   for multi-line function calls, and other small fixes (tickets #38 and #39).
 
 
 Version 0.8.1