Revision history for Perl extension XML::Parser.
2.33
- Fixed Tree style (grantm)
- Fixed some non-utf8 stuff in DTDs (patch in XML::DOM tarball)
2.32
- Memory leak fix (Juerd Waalboer).
- Added windows-1252 encoding
- Styles moved to separate .pm files to make loading faster and
ease maintainence
- Don't load IO::Handle unless we really need to
2.31 Tue Apr 2 13:39:51 EST 2002
- Ilya Zakharevich <ilya@math.ohio-state.edu> and
Dave Mitchell <davem@fdgroup.com> both provided patches to
fix problems module had with 5.8.0
- Dave Mitchell also made some UTF-8 related fixes to the test suite.
2.30 Thu Oct 5 12:47:36 EDT 2000
- Get rid of ContentStash global. Not that big a deal looking it up
everytime and gets rid of a potential threading problem.
- Switch to shareable library version of expat from sourceforge
(i.e. no longer include expat source and require that libexpat
be installed)
- Bob Tribit <btribit@traffic.com> demonstrated a fix for problems
in compiling under perl 5.6.0 with 5.005 threading.
- Matt Sergeant <matt@sergeant.org> discovered a typo ('IO::Handler'
instead of 'IO::Handle') in Expat.pm that caused IO::Handle objects
to be treated as strings instead of handles.
- Matt Sergeant also provided a patch to allow tied handles to work
properly in calls to parse.
- Eric Bohlman <ebohlman@netcom.com> reported a failure when
incremental parsing and external parsing were used together.
Need to give explicit package when calling Do_External_Parse
from externalEntityRef otherwise fails when called through ExpatNB.
2.29 Sun May 21 21:19:45 EDT 2000
- In expat, notation declaration handler registration wasn't
surviving through external entity references.
- Chase Tingley <tingley@sundell.net> discovered that text
accumulation in the Stream style wasn't working across processing
instructions and recommended the appropriate fix.
- Jochen Wiedmann <jochen.wiedmann@softwareag.com>, noted that
you couldn't use ExpatNB directly because it wasn't setting
the protective _State_ variable. Now doing this in the
parse_more method of ExpatNB.
- At the suggestion of Grant Hopwood <hopwoodg@valero.com>, now
calling the env_proxy method on the LWP::UserAgent in the LWP
external entity handler when it's created to set any proxies
from environment variables.
- Grant McLean, Matt Sergeant (& others I may have missed) noted that
loading the LWP & URI modules slowed startup of the module, even
if the application didn't need it. The default LWP handler is now
dynamicly loaded (along with LWP & URI modules) the first time an
external entity is referenced. Also provided a NoLWP option to
XML::Parser that forces the file based external entity handler.
- Fixed allocation errors in element declaration patches in expat
- The Expat base method now works, even before expat starts parsing.
- Changed the canonical script to take an optional file argument.
- Enno Derksen <enno@att.com> reported that the attlist handler
was not returning NOTATION type attlist information.
- Michel Rodriguez <mrodrigu@ieee.org>, noted that the constructor
for XML::Parser objects no longer checked for the existence of
applications installed external entity handlers before installing
the default ones.
- Burkhard Meier <burkhard.meier@ixos.de> sent in a fix for
compiler directives in Expat/Makefile.PL for Win32 machines.
A change in 5.6.0 caused the old conditional to fail.
- Forgot to document changes to the Entity declaration handler:
there is an additional "IsParam" argument that indicates whether
or not the entity is a parameter entity. This information is
no longer passed on in the name.
- Ben Low <ben@snrc.uow.edu.au> reported an undefined macro with
version 5.004_04.
2.28 Mon Mar 27 21:21:50 EST 2000
- Junked local (Expat.xs) declaration parsing and patched expat to
handle XML declarations, element declarations, attlist declarations,
and all entity declarations. By eliminating both shadow buffers and
local declaration parsing in Expat.xs, I've eliminated the two most
common sources of serious bugs in the expat interface.
o thus fixed the segfault and parse position bugs reported by
Ivan Kurmanov <iku@fnmail.com>
o and the doctype bug reported by Kevin Lund
<Kevin.Lund@westgroup.com>
o The element declaration handler no longer receives a string,
but an XML::Parser::ContentModel object that represents the
parsed model, but still looks like a string if referred to as
a string. This class is documented in the XML::Parser::Expat
pod under "XML::Parser::ContentModel Methods".
o The doctype declaration handler no longer receives the internal
subset as a string, but in its place a true or undef value
indicating whether or not there is an internal subset. Also,
it's called prior to processing either the internal or external
DTD subset (as suggested by Enno Derksen <enno@att.com>.)
o There is a new DoctypeFin handler that's called after finishing
parsing all of the DOCTYPE declaration, including any internal
or external DTD declarations.
o One bit of lossage is that recognized_string, original_string,
and default_current no longer work inside declaration handlers.
- Added a handler that gets called after parsing external entities:
ExternEntFin. Suggested by Jeff Horner <jhorner@netcentral.net>.
- parsefile, file_ext_ent_handler, & lwp_ext_ent_handler now all
set the base path. This problem has been raised more than once
and I'm not sure to whom credit should be given.
- The file_ext_ent_handler now opens a file handle instead of
reading the entire entity at once.
- Merged patches supplied by Larry Wall to (for perl 5.6 and beyond)
tag generated strings as UTF-8, where appropriate.
- Fixed a bug in xml_escape reported by Jerry Geiger <jgeiger@rios.de>.
It failed when requesting escaping of perl regex meta-characters.
- Laurent Caprani <caprani@pop.multimania.com> reported a bug in the
Proc handler for the Debug style.
- <chocolateboy@usa.net> sent in a patch for the element index
mechanism. I was popping the stack too soon in the endElement fcn.
- Jim Miner <jfm@winternet.com> sent in a patch to fix a warning in
Expat.pm.
- Kurt Starsinic pointed out that the eval used to check for string
versus IO handle was leaving $@ dirty, thereby foiling higher
level exception handlers
- An expat question by Paul Prescod <paul@prescod.net> helped me
see that exeptions in the parse call bypass the Expat release method,
causing memory leaks.
- Mark D. Anderson <mda@discerning.com> noted that calling
recognized_string from the Final method caused a dump. There are
a bunch of methods that should not be called after parsing has
finished. These now have protective if statements around them.
- Updated canonical utility to conform to newer version of Canonical
XML working draft.
2.27 Sat Sep 25 18:26:44 EDT 1999
- Corrected documentation in Parser.pm
- Deal with XML_NS and XML_BYTE_ORDER macros in Expat/Makefile.PL
- Chris Thorman <chris@thorman.com> noted that "require 'URI::URL.pm'"
in Parser.pm was in error (should be "require 'URI/URL.pm'")
- Andrew McNaughton <andrew@scoop.co.nz> noted "use English" and
use of '$&' slowed down regex handling for whole application, so
they were excised from XML::Parser::Expat.
- Work around "modification of read-only value" bug in perl 5.004
- Enno Derksen <enno@att.com> reported that the Doctype handler
wasn't being called when ParseParamEnt was set.
- Now using Version 19990728 of expat, with local patches.
- Got rid of shadow buffer
o thus fixed the error reported by Ashley Sanders
<a.sanders@mcc.ac.uk>
o and removed ExpatNB limitations that Peter Billam
<music@pjb.com.au> noted.
- Vadim Konovalov <vkonovalov@lucent.com> had a problem compiling
for multi-threading that was fixed by changing Perl_sv_setsv to
sv_setsv.
- Added new Expat method: skip_until(index)
- Backward incompatible change to method xml_escape: to get former
behavior use $xp->xml_escape($string, '>', ...)
- Added utility, canonical, to samples
2.26 Sun Jul 25 19:06:41 EDT 1999
- Ken Beesley <ken.beesley@xrce.xerox.com> discovered that
declarations in the external subset are not sent to registered
handlers when there is no internal subset.
- Fixed parse_dtd to work when entity values or attribute defaults
are so large that they might be broken across multiple calls to
the default handler.
- For lwp_ext_ent_handler, use URI::URL instead of URI so that old
5.004 installations will work with it.
2.25 Fri Jul 23 06:23:43 EDT 1999
- Now using Version 1990709 of expat. No local patches.
- Numerous people reported a SEGV problem when running t/cdata
on various platforms and versions of perl. The problem was
introduced with the setHandlers change. In some cases an
un-initialized value was being returned.
- Added an additional external entity handler, lwp_ext_ent_handler,
that deals with general URIs. It is installed instead of the
"file only" handler if the LWP package is installed.
2.24 Thu Jul 8 23:05:50 EDT 1999
- KangChan Lee <dolphin@comeng.chungnam.ac.kr> supplied the
EUC-KR encoding map.
- Enno Derksen <enno@att.com> forwarded reports by Jon Eisenzopf
<eisen@pobox.com> and Stefaan Onderbeke <onderbes@bec.bel.alcatel.be>
about a core dump using XML::DOM. This was due to a bug in the
prolog parsing part of XML::Parser.
- Loic Dachary <loic@ceic.com> discovered that changing G_DISCARD to
G_VOID introduced a small memory leak. Changed G_VOID back to
G_DISCARD.
- As suggested by Ben Holzman <bholzman@earthlink.net>, the
setHandlers methods of both Parser and Expat now return lists that
consist of type, handler pairs that correspond to the input, but
the handlers returned are the ones that were in effect prior to
the call.
- Now using Version 19990626 of expat with a local patch (provided
by James Clark.)
- Added option ParseParamEnt. When set to a true value, parameter
entities are parsed and the external DTD is read (unless standalone
set to "Yes" in document).
2.23 Mon Apr 26 21:30:28 EDT 1999
- Fixed a bug in the ExpatNB class reported by Gabe Beged-Dov
<begeddov@jfinity.com>. The ErrorMessage attribute wasn't
being initialized for ExpatNB. This should have been done in
the Expat constructor.
- Applied patch provided by Nathan Kurz <nate@valleytel.net> to
fix more perl stack manipulation errors in Expat.xs.
- Applied another patch by Nathan to change perl_call_sv flag
from G_DISCARD to G_VOID for callbacks, which helps performance.
- Murata Makoto <murata@apsdc.ksp.fujixerox.co.jp> reported a
problem on Win32 platforms that only showed up when UTF-16 was
being used. The needed call to binmode was added to the parsefile
methods.
- Added documentation for release method that was added in release
2.20 to Expat pod. (Point raised by <mookie@undef.com>)
- Now using Version 19990425 of expat. No local patches.
- Added specified_attr method and made ineffective the is_defaulted
method.
2.22 Sun Apr 4 11:47:25 EDT 1999
- Loic Dachary <loic@ceic.com> reported a core dump with a small
file with a comment that wasn't properly closed. Fixed in expat
by updating positionPtr properly in final call of XML_Parse.
(Reported to & acknowledged by James Clark.)
- Made more fixes to Expat.xs position calculation.
- Loic Dachary <loic@ceic.com> provided patches for fixing a
memory growth problem with large documents. (Garbage collection
wasn't happening frequently enough.)
- As suggested by Gabe Beged-Dov <begeddov@jfinity.com>, added
a non-blocking parse mechanism:
- Added parse_start method to XML::Parser, which returns a
XML::Parser::ExpatNB object.
- Added XML::Parser::ExpatNB class, which is a subclass of
Expat and has the additional methods parse_more & parse_done
- Made some performance tweaks as suggested by performance thread
on perl-xml discussion list. [With negligible results]
- Tried to clarify Tree style structure in Parser pod
2.21 Sun Mar 21 17:42:04 EST 1999
- Warren Vik <wvik@whitebarn.com> provided patches for a bug
introduced with the is_defaulted method. It manifested itself
by bogusly reporting duplicate attributes.
- Now using latest expat from ftp://ftp.jclark.com/pub/test/expat.zip,
Version 19990307. (Plus any patches in Expat/expat.patches.)
- As suggested by Tim Bray, added an xml_escape method to
Expat.
- Murray Nesbitt <murray@activestate.com> had build problems
on Win32 that were solved by swapping 2 include files in
Expat.xs
- Added following Expat namespace methods:
new_ns_prefixes
expand_ns_prefix
current_ns_prefixes
- Fixed memory handling in recognized_string method to get rid
of "Attempt to free unreferenced scalar" bug.
2.20 Sun Feb 28 15:35:52 EST 1999
- Fixed miscellaneous bugs in xmlfilter.
- In the default external entity handler, prepend the base only
for relative URLs.
- Chris Nandor <pudge@pobox.com> provided patches for building
on Macintosh.
- As suggested by Matt Sergeant <Matthew.Sergeant@eml.ericsson.se>,
added the finish method to Expat.
- Matt also provided a fix to a bug he discovered in the Streams
style.
- Fixed a parse position bug reported by Enno Derksen <enno@att.com>
that was affecting both original_string and position_in_context.
- Fixed a gross memory leak reported by David Megginson,
<david@megginson.com>: there was a circular reference to the Expat
object and the internal end handler for context was not freeing
element names after they were removed from the context stack.
- Now using expat Version 19990109
(Plus any patches in Expat/expat.patches)
- Added is_defaulted method to Expat to tell if an attribute
was defaulted. (Requested by Enno Derksen for XML::DOM.)
- Matt Sergeant <Matthew.Sergeant@eml.ericcson.se> reported that
the XML::Parser parse methods weren't propagating array context
to the Final handler. Now they are.
- Fixed more memory leaks (again reported by David Megginson).
The SVs pointing to the handlers weren't being reclaimed when
the callback vector was freed.
- Added the element_index method to Expat.
2.19 Sun Jan 3 11:23:45 EST 1999
- When the recognized string is long enough, expat uses multiple
calls to reportDefault. Fixed recString handler in Expat.xs to
deal with this properly.
- Added original_string method to Expat. This returns the untranslated
string (i.e. original encoding) that caused current event.
- Alberto Accomazzi <alberto@cfa0.harvard.edu> sent in more patches
for perl5.005_54 incompatibilities.
- Alberto also fingered a nasty memory bug in Expat.xs that arose
sometimes when you registered a declaration handler but no
default handler. It would give you a "Not a CODE reference"
error in a place that wasn't using any CODE references.
- <schinder@pobox.com> reported a problem with compiling expat
on a Sun 4 due to non-exsitance of memmove on that OS. Provided
a workaround in Makefile.PL
- Now using expat Version 19981231 from James Clark's test directory.
- Made patch to this version in order to support original_string
(see Expat/expat.patches.)
- Added CdataStart and CdataEnd handlers to expat.
2.18 Sun Dec 27 07:39:23 EST 1998
- Alberto Accomazzi <alberto@cfa0.harvard.edu> pointed out that
the DESTROY sub in the new XML::Parser::Encinfo package was
pointing to the wrong package for calling FreeEncoding.
- Tarang Kumar Patel <mombasa@ptolemy.arc.nasa.gov> reported
the mis-declaration of an integer as unsigned in the
convert_to_unicode function in Expat.xs.
- Glenn R. Kronschnabl <grk@arlut.utexas.edu> reported a problem
with ExternEnt handlers when using parsefile. Turned out to be
an unmatched ENTER; SAVETMPS pair that screwed up the Perl stack.
- Tom Hughes <tom@compton.demon.co.uk> reported that the fix I put
in for the swith to PL_sv.. names failed with 5.0005_54, since
these became real variables instead of macros. Switched to just
checking the PATCHLEVEL macro.
- Yoshida Masato <yoshidam@inse.co.jp> provided the EUC-JP encodings
(the corresponding XML files are in XML::Encoding 1.01 or later.)
- With the advice of MURATA Makoto <murata@apsdc.ksp.fujixerox.co.jp>,
removed the Shift_JIS encoding and replaced it with 4 variations
he provided. He also provided an explanatory message.
- Added the recognized_string method to Expat, deprecating
default_current.
- Now using expat Version 19981122 from James Clark's test directory
(this fixes another bug with external entity reference handlers)
- Added a default external entity handler that only accesses file:
based URLs.
2.17 Sun Dec 13 17:39:58 EST 1998
- Replaced uses of malloc, realloc, and free with New, Renew,
and Safefree respectively
- In Expat.pm, fixed methods in_element and within_element to
work correctly with namespaces.
- xmlfilter - Substitute quoted equivalents for special characters
in attribute values.
- position_in_context was off by one line when position was at
the end of line.
- For the context methods in Expat.pm, do the right thing when
the context list is empty.
- Added methods xpcroak and xpcarp to Expat.
- Alberto Accomazzi <alberto@cfa0.harvard.edu> noted that perl
releases 5.005_5* (the pre 5.006 development versions) won't
accept sv_undef (and related constants) anymore and we have
to switch to PL_sv_...
- Alberto also reported a warning in the newer versions of
IO::Handle about input_record_separator not being treated on
a per-handle basis.
- Fixed bug that Jon Udell <udell@top.monad.net> reported in
Stream style: Text handler most of the time didn't see proper
context.
- Added XML::Parser::Expat::load_encoding function and support
for external encodings.
2.16 Tue Oct 27 22:27:33 EST 1998
- Fixed bug reported by Enno Derksen <enno@att.com>:
Now treats parameter entity declarations correctly. The entity
handler sees the name beginning with '%' if it's a parameter
entity declaration.
- Nigel Hutchison <nwoh@software-ag.de> pointed out that stream.t
wasn't portable off Unix systems. Replaced with portable version.
- Fixed bug reported by Enno Derksen <enno@att.com>:
XML Declaration was firing off both XMLDecl handler *and* Default
handler.
- Added option NoExpand to Expat to turn off expansion of entity
references when a default handler is set.
2.15 Tue Oct 20 14:50:11 EDT 1998
- In Expat's parse method, account for undefined previous
record separators.
- Simplify a couple of Expat methods.
- Re-ordered Changes entries to put latest changes first.
- In XML::Parser::new, set Handlers if not already set
- New Handler (XMLDecl) for handling XML declarations
- New Handler (Doctype) for handling DOCTYPE declarations
- New Handler (Entity) for handling ENTITY declarations in
the internal subset.
- New Handler (Element) for handling ELEMENT declarations in
the internal subset.
- New Handler (Attlist) for handling ATTLIST declarations in
the internal subset.
- Documented new handlers
- Added t/decl.t to test new handlers
2.14 Sun Oct 11 22:17:15 EDT 1998
- Always use method calls for streams.
- Use perl's input_record_separator to find delimiter (i.e. each
"line" is an entire XML doc with delimiter appended)
- Deal with line being longer than buffer.
2.13 Thu Oct 8 16:58:39 EDT 1998
- Fixed a major oops in Expat.xs where I was trying to decrement
a refcnt on an unallocated SV, leading to a segment violation.
(Why did this show up on HPUX but not Linux?)
2.12 Thu Oct 8 00:05:10 EDT 1998
- Incorporated fix to t/astress.t from <fletch@phydeaux.org> (Mike
Fletcher).
- Change to xmlstats from <dblack@candle.superlink.net> (David
Alan Black)
- Access Handlers_Setters in Expat and Handler_Types in Parser
through object reference (following admonition in perltoot
about class data.)
- Added Stream_Delimiter option to Expat.
- In the parse_stream function in Expat.xs, if we either have a
Stream_Delimiter or if there's no file descriptor, use method
calls instead. For Stream_Delimiter in particular, the function
now uses the getline method so it can check for the delimiter
without consuming stuff past the delimiter from the stream.
2.11 Sun Oct 4 22:15:53 EDT 1998
- Swapped out local patch for expat and swapped in James Clark's
patch.
- Pass on all Parser attributes (other than those excluded by
Non_Expat_Options) to the instance of Expat created at parse time.
- New method for Expat: generate_ns_name
- Split test.pl into t/*.t and change Makefile.PL so we don't do a
useless descent into Expat subdir for testing.
- Stop the numeric warning for eq_name and namespace method.
2.10 Fri Sep 25 18:36:46 EDT 1998
- Uses expat Version 19980924
(with local patch - see Expat/expat/xmlparse/xmlparse.c.diff)
- Use newSVpvn when PERL_VERSION >= 5.005
- Completed xmlfilter
- Added support for namespace processing:
o Namespaces option to XML::Parser and XML::Parser::Expat
o Two new methods in Expat:
namespace - to return namespace associated with name
eq_name - compare 2 names for equality across namespaces.
- Use expat's new SetDefaultHandlerExpand instead of SetDefaultHandler
so that entity expansion may continue even if the default handler
is set.
- Moved test.pl back up main level and changed to work with XML::Parser
- Added tests for namespaces
2.09 Fri Sep 18 10:33:38 EDT 1998
- Fixed errors that caused -w to fret in XML::Parser.
- Fixed depth method in XML::Parser::Expat
- There were a few places in Expat.xs where garbage strings may
have been returned due to the expat library giving us zero-length
strings. Fixed by using a local version of newSVpv where length
means length, even when zero.
- The default handler setter in Expat.xs, was inappropriately setting
cbv->dflt_sv when there was a null handler.
2.08 Thu Sep 17 11:47:13 EDT 1998
- Make XML::Parser higher-level re-usable parser objects. Old object
now becomes XML::Parser::Expat.
- The XML::Parser object now supports the style mechanism very close
to that in the 1.0 version.
2.07 Wed Sep 9 11:03:43 EDT 1998
- Added some samples (xmlcomments & xmlstats)
- Now requires 5.004 (due to sv_catpvf)
- Changed Makefile.PL to allow automatic manification
- Added a test that reads xml spec (to check buffer boundary errors)
2.06 Tue Sep 1 10:40:41 EDT 1998
- Fixed the methods current_line, current_byte, and current_column
- Added some tests
2.05 Mon Aug 31 15:29:42 EDT 1998
- Made Makefile.PL changes suggested by Murray Nesbitt
<murray@ActiveState.com> to support building on Win32
and for making PPM binaries.
- Added method parse
- Changed parsestring and parsefile to use new parse method
- Deprecated parsestring method
- Improved error handling in the ExternEnt handler
2.04 Wed Aug 26 13:25:01 EDT 1998
- Uses expat Version 1.0 of August 14, 1998
- Some document changes
- Changed dist section in Makefile.PL
- Added ExternEnt handler
- Added tests for ExternEnt
2.03 Fri Aug 21 17:19:26 EDT 1998
- Changed InitEncoding to ProtocolEncoding. Default to none.
Pass null string to expat's ParserCreate when there is no
ProtocolEncoding.
- Fixed bug in parsefile & parsestring where they were referring
to an ErrorContext *method* instead of a field.
- Fixed position_in_context bugs:
-- 'last' in do {} while ();
-- insert newline before pointer when no following newline
in buffer.
- Added some additional tests
2.02 Thu Aug 20 14:05:08 EDT 1998
- Fixed parsefile problem reported by
"Robert Hanson" <robertha@zenweb.com>, using a modification of
his suggested fix.
- Responded to problem reported by
Bart Schuller <schuller+perl-xml@lunatech.com>
by pre-expanding parts of the XML_UPD macro to avoid confusing
some versions of gcc.
- Changed the constructor to take the option InitEncoding, which
gets passed to the ParserCreate call. When not given, defaults
to UTF-8.
- Added method position_in_context
- Added Constructor option ErrorContext and added reporting of
errors in context.
2.01 Wed Aug 19 11:42:42 EDT 1998
- Added methods:
default_current, base, current_line, current_column,
current_byte, context
- Added some tests
- parsestring and parsefile now croak if they're re-used
- Filled in some documentation
2.00 Mon Aug 17 12:01:33 EDT 1998
- repackaged with James Clark's most recent expat
- changed to an API closer to expat
1.00 March 1998
- Larry Wall's original version