Change history for HTML-Parser

3.81      2023-01-30
    - Stop depending on "Test" (GH#34) (James Raspass)
    - fix test scripts after conversion to Test::More (GH#35) (Graham Knop)

3.80      2022-11-01
  * Fix compatibility with ancient perl by avoiding index in test (GH#33)
    (Graham Knop)

3.79      2022-10-12
  * Modernise XS (GH#32) (James Raspass)
  * Skip threads on older perl versions, as they often segfault (GH#31) (Graham
  * Knop)

3.78      2022-03-28
  * Remove unused variable (GH#26) (Michal Josef Špaček)

3.77      2022-03-14
  * Update tests to remove HTML4 specific tags (GH#25) (Jess)

3.76      2021-03-04
  * Add a fix for a stack confusion error on `eof`. (GH#21) (Matthew Horsfall
    and Chase Whitener)

3.75      2020-08-30
  * Clean up the prereqs a bit
  * Mark HTML::Filter as deprecated as the docs point out
  * Move into the lib directory with the others. This will help
    with everything from auto version bumps after releases, to scanning for
    prerequisites and spelling errors.
  * Fix a few spelling errors in the POD for HTML::Parser
  * Clean up the spacing on many examples in HTML::Parser

3.74      2020-08-30
  * Fix the order of date and version in this change log. (Thanks, haarg)
  * Convert to Dist::Zilla
  * Build all prereqs from our cpanfile
  * Go through all test files and:
    * perltidy
    * Use strict/warnings
    * Get rid of two-arg open
    * Get rid of BAREWORD filehandles
    * Fix the eval pattern used
    * Only use -w where we catch $SIG{__WARN__}
    * Fix encoding problems
    * use utf8 where we have unicode in the source
  * Fix a typo here and there
  * perltidy all of the example apps in eg/
  * Add comments explaining the apps in eg/ (GH#13 Thanks, Salvatore Bonaccorso)
  * Print out UTF-8 encoded data where sensible in eg/

3.73     2020-08-24
  * Cleaned up this changes log.
  * Added a .mailmap file to organize contributions accurately.
  * Ensure all versions are equal and on the current version
  * Add the .mailmap to the MANIFEST
  * Change the META information to point to the new GH repository
  * Add a .perltidyrc to use going forward
  * Add hctype.h and pfunc.h to the dist as static files and stop asking
    for them to be built on the user's end.
  * Remove t/pod.t from userland testing
  * Remove t/pod-coverage.t from userland testing
  * Clean up the MANIFEST
  * Start testing via GitHub Actions/Workflows
  * Protect active parser from being freed (PR 13, RT #115034)

3.72     2016-01-19
  * Avoid more clang casting warnings
  * Remove trailing whitespace
  * Ensure entities expand to utf8 sequences under 'utf8_mode' [RT#99755]
  * typo fixes (David Steinbrunner)
  * Silence clang warning (Jacques Germishuys)
  * const+static-ing (bulk88)

3.71     2013-05-09
  * Transform ':' in headers to '-' [RT#80524]

3.70     2013-03-28
  * Fix for cross-compiling with Buildroot (François Perrad)
  * Comment typo fix
  * Fix Issue #3 / RT #84144: HTML::Entities::decode_entities() needs
    to call SV_CHECK_THINKFIRST() before checking READONLY flag (Yves Orton)

3.69     2011-10-15
  * Documentation fix; encode_utf8 mixup [RT#71151]
  * Make it clearer that there are 2 (actually 3) options for handing "UTF-8 garbage"
  * Github is the official repo
  * Can't be bothered to try to fix the failures that occur on perl-5.6
  * fix to TokeParser to correctly handle option configuration (Barbie)
  * Aesthetic change: remove extra ; (Jon Jensen)
  * Trim surrounding whitespace from extracted URLs. (Ville Skyttä)

3.68     2010-09-01
  * Declare the encoding of the POD to be utf8

3.67     2010-08-17
  * bleadperl 2154eca7 breaks HTML::Parser 3.66 [RT#60368] (Nicholas Clark)

3.66     2010-07-09
  * Fix entity decoding in utf8_mode for the title header

3.65     2010-04-04
  * Eliminate buggy entities_decode_old
  * Fixed endianness typo [RT#50811] (Salvatore Bonaccorso)
  * Documentation Fixes. (Ville Skyttä)

3.64     2009-10-25
  * Convert files to UTF-8
  * Don't allow decode_entities() to generate illegal Unicode chars
  * Copyright 2009
  * Remove rendundant (repeated) test
  * Make parse_file() method use 3-arg open [RT#49434]

3.63     2009-10-22
  * Take more care to prepare the char range for encode_entities [RT#50170]
  * decode_entities confused by trailing incomplete entity

3.62     2009-08-13
  * Doc patch: Make it clearer what the return value from ->parse is
  * HTTP::Header doc typo fix. (Ville Skyttä)
  * Do not bother tracking style or script, they're ignored. (Ville Skyttä)
  * Bring HTML 5 head elements up to date with WD-html5-20090423. (Ville Skyttä)
  * Improve HeadParser performance. (Ville Skyttä)

3.61     2009-06-20
  * Test that triggers the crash that Chip fixed
  * Complete documented list of literal tags
  * Avoid crash (referenced pend_text instead of skipped_text) (Chip Salzenberg)
  * Reference HTML::LinkExttor [RT#43164] (Antonio Radici)

3.60     2009-02-09
  * Spelling fixes. (Ville Skyttä)
  * Test multi-value headers. (Ville Skyttä)
  * Documentation improvements. (Ville Skyttä)
  * Do not terminate head parsing on the <object> element (added in HTML 4.0). (Ville Skyttä)
  * Add support for HTML 5 <meta charset> and new HEAD elements. (Ville Skyttä)
  * Short description of the htextsub example (Damyan Ivanov)
  * Suppress warning when encode_entities is called with undef [RT#27567] (Mike South)
  * HTML::Parser doesn't compile with perl 5.8.0. (Zefram)

3.59     2008-11-24
  * Restore perl-5.6 compatibility for HTML::HeadParser.
  * Improved META.yml

3.58     2008-11-17
  * Suppress "Parsing of undecoded UTF-8 will give garbage" warning
     with attr_encoded [RT#29089]
  * HTML::HeadParser:
       - Recognize the Unicode BOM in utf8_mode as well [RT#27522]
       - Avoid ending up with '/' keys attribute in Link headers.

3.57     2008-11-16
  * The <iframe> element content is now parsed in literal mode.
  * Parsing of <script> and <style> content ends on the first end tag
     even when that tag was in a quoted string.  That seems to be the
     behaviour of all modern browsers.
  * Implement backquote() attribute as requested by Alex Kapranoff.
  * Test and documentation tweaks from Alex Kapranoff.

3.56     2007-01-12
  * Cloning of parser state for compatibility with threads.
     Fixed by Bo Lindbergh <>.
  * Don't require whitespace between declaration tokens.

3.55     2006-07-10
  * Treat <> at the end of document as text.  Used to be
     reported as a comment.
  * Improved Firefox compatibility for bad HTML:
      - Unclosed <script>, <style> are now treated as empty tags.
      - Unclosed <textarea>, <xmp> and <plaintext> treat rest as text.
      - Unclosed <title> closes at next tag.
  * Make <!a'b> a comment by itself.

3.54     2006-04-28
  * Yaakov Belch discovered yet another issue with <script> parsing.
     Enabling of 'empty_element_tags' got the parser confused
     if it found such a tag for elements that are normally parsed
     in literal mode.  Of these <script src="..."/> is the only
     one likely to be found in documents.

3.53     2006-04-27
  * When ignore_element was enabled it got confused if the
     corresponding tags did not nest properly; the end tag
     was treated it as if it was a start tag.
     Found and fixed by Yaakov Belch <>.

3.52     2006-04-26
  * Make sure the 'start_document' fires exactly once for
     each document parsed.  For earlier releases it did not
     fire at all for empty documents and could fire multiple
     times if parse was called with empty chunks.
  * Documentation tweaks and typo fixes.

3.51     2006-03-22
  * Named entities outside the Latin-1 range are now only expanded
     when properly terminated with ";".  This makes HTML::Parser
     compatible with Firefox/Konqueror/MSIE when it comes to how these
     entities are expanded in attribute values.  Firefox does expand
     unterminated non-Latin-1 entities in plain text, so here
     HTML::Parser only stays compatible with Konqueror/MSIE.
     Fixes <>.
  * Fixed some documentation typos spotted by <>.

3.50     2006-02-14
  * The 3.49 release didn't compile with VC++ because it mixed code
     and declarations.  Fixed by Steve Hay <>.

3.49     2006-02-08
  * Events could sometimes still fire after a handler has signaled eof.
  * Marked_sections with text ending in square bracket parsed wrong.
     Fix provided by <>.

3.48     2005-12-02
  * Enabling empty_element_tags by default for HTML::TokeParser
     was a mistake.  Reverted that change.
  * When processing a document with "marked_sections => 1", the
     skipped text missed the first 3 bytes "<![".

3.47     2005-11-22
  * Added empty_element_tags and xml_pic configuration
     options.  These make it possible to enable these XML
     features without enabling the full XML-mode.
  * The empty_element_tags is enabled by default for

3.46     2005-10-24
  * Don't try to treat an literal &nbsp; as space.
     This breaks Unicode parsing.
  * The unbroken_text option is now on by default
     for HTML::TokeParser.
  * HTML::Entities::encode will now encode "'" by default.
  * Improved report/ignore_tags documentation by
     Norbert Kiesel <>.
  * Test suite now use Test::More, by
     Norbert Kiesel <>.
  * Fix HTML::Entities typo spotted by
     Stefan Funke <>.
  * Faster load time with XSLoader (perl-5.6 or better now required).
  * Fixed POD markup errors in some of the modules.

3.45     2005-01-06
  * Fix stack memory leak caused by missing PUTBACK.  Only
     code that used $p->parse(\&cb) form was affected.
     Fix provided by Gurusamy Sarathy <>.

3.44     2004-12-28
  * Fix confusion about nested quotes in <script> and <style> text.

3.43     2004-12-06
  * The SvUTF8 flag was not propagated correctly when replacing
     unterminated entities.
  * Fixed test failure because of missing binmode on Windows.

3.42     2004-12-04
  * Avoid sv_catpvn_utf8_upgrade() as that macro was not
     available in perl-5.8.0.
     Patch by Reed Russell <>.
  * Add casts to suppress compilation warnings for char/U8
  * HTML::HeadParser will always push new header values.
     This make sure we never loose old header values.

3.41     2004-11-30
  * Fix unresolved symbol error with perl-5.005.

3.40     2004-11-29
  * Make utf8_mode only available on perl-5.8 or better.  It produced
     garbage with older versions of perl.
  * Emit warning if entities are decoded and something in the first
     chunk looks like hi-bit UTF-8.  Previously this warning was only
     triggered for documents with BOM.

3.39_92     2004-11-23
  * More documentation of the Unicode issues.  Moved around HTML::Parser
     documentation a bit.
  * New boolean option; $p->utf8_mode to allow parsing of raw  UTF-8.
  * Documented that HTML::Entities::decode_entities() can take multiple
  * Unterminated entities are now decoded in text (compatibility
     with MSIE misfeature).
  * Document HTML::Entities::_decode_entities(); this variation of the
     decode_entities() function has been available for a long time, but
     have not been documented until now.
  * HTML::Entities::_decode_entities() can now be told to try to
     expand unterminated entities.
  * Simplified Makefile.PL

3.39_91     2004-11-23
  * The HTML::HeadParser will skip Unicode BOM.  Previously it
     would consider the <head> section done when it saw the BOM.
  * The parser will look for Unicode BOM and give appropriate
     warnings if the form found indicate trouble.
  * If no matching end tag is found for <script>, <style>, <xmp>
     <title>, <textarea> then generate one where the next tag
  * For <script> and <style> recognize quoted strings and don't
     consider end element if the corresponding end tag is found
     inside such a string.

3.39_90     2004-11-17
  * The <title> element is now parsed in literal mode, which
     means that other tags are not recognized until </title> has
     been seen.
  * Unicode support for perl-5.8 and better.
  * Decoding Unicode entities always enabled; no longer a compile
    time option.
  * Propagation of UTF8 state on strings.
    Patch contributed by John Gardiner Myers <>.
  * Calculate offsets and lengths in chars for Unicode strings.
  * Fixed link typo in the HTML::TokeParser documentation.

3.38     2004-11-11
  * New boolean option; $p->closing_plaintext
     Contributed by Alex Kapranoff <>

3.37     2004-11-10
  * Improved handling of HTML encoded surrogate pairs and illegally
     encoded Unicode; <>.
     Patch by John Gardiner Myers <>.
  * Avoid generating bad UTF8 strings when decoding entities
     representing chars beyond #255 in 8-bit strings.  Such bad
     UTF8 sometimes made perl-5.8.5 and older segfault.
  * Undocument v2 style subclassing in synopsis section.
  * Internal cleanup: Make 'gcc -Wall' happier.
  * Avoid modification of PVs during parsing of attrspec.
    Another patch by John Gardiner Myers.

3.36     2004-04-01
  * Improved MSIE/Mozilla compatibility.  If the same attribute
     name repeats for a start tag, use the first value instead
     of the last.  Patch by Nick Duffek <>.

3.35     2003-12-12
  * Documentation fixes by Paul Croome <>.
  * Removed redundant dSP.

3.34     2003-10-27
  * Fix segfault that happened when the parse callback caused
     the stack to get reallocated.  The original bug report was

3.33     2003-10-14
  * Perl 5.005 or better is now required.  For some reason we get
     a test failure with perl-5.004 and I don't really feel like
     debugging that perl any more.  Details about this failure can
     be found at <>.
  * New HTML::TokeParser method called 'get_phrase'.  It returns
     all current text while ignoring any phrase-level markup.
  * The HTML::TokeParser method 'get_text' now expands skipped
     non-phrase-level tags as a single space.

3.32     2003-10-10
  * If the document parsed ended with some kind of unterminated markup,
     then the parser state was not reset properly and this piece of markup
     would show up in the beginning of the next document parsed.
  * The get_text and get_trimmed_text methods of HTML::TokeParser can
     now take multiple end tags as argument.  Patch by <>
     at <>.
  * Various documentation tweaks.
  * Included another example program: hdump

3.31     2003-08-19
  * The -DDEBUGGING fix in 3.30 was not really there :-(

3.30     2003-08-17
  * The previous release failed to compile on a -DDEBUGGING perl
     like the one provided by Redhat 9.
  * Got rid of references to perl-5.7.
  * Further fixes to avoid warnings from Visual C.
     Patch by Steve Hay <>.

3.29     2003-08-14
  * Setting xml_mode now implies strict_names also for end tags.
  * Avoid warning from Visual C.  Patch by <>.
  * 64-bit fix from Doug Larrick <>
  * Try to parse similar to Mozilla/MSIE in certain edge cases.
     All these are outside of the official definition of HTML but
     HTML spam often tries to take advantage of these.
       - New configuration attribute 'strict_end'.  Unless enabled
         we will allow end tags to contain extra words or stuff
         that look like attributes before the '>'.  This means that
         tags like these:
            </foo foo="<ignored>">
            </foo ignored>
            </foo ">" ignored>
         are now all parsed as a 'foo' end tag instead of text.
         Even if the extra stuff looks like attributes they will not
         be reported if requested via the 'attr' or 'tokens' argspecs
         for the 'end' handler.
       - Parse '</:comment>' and '</ comment>' as comments unless
         strict_comment is enabled.  Previous versions of the parser
         would report these as text.  If these comments contain
         quoted words prefixed by space or '=' these words can
         contain '>' without terminating the comment.
       - Parse '<! "<>" foo>' as comment containing ' "<>" foo'.
         Previous versions of the parser would terminate the comment
         at the first '>' and report the rest as text.
       - Legacy comment mode:  Parse with comments terminated with a
         lone '>' if no '-->' is found before eof.
       - Incomplete tag at eof is reported as a 'comment' instead
         of 'text' unless strict_comment is enabled.

3.28     2003-04-16
  * When 'strict_comment' is off (which it is by default)
     treat anything that matches <!...> a comment.
  * Should now be more efficient on threaded perls.

3.27     2003-01-18
  * Typo fixes to the documentation.
  * HTML::Entities::escape_entities_numeric contributed
     by Sean M. Burke <>.
  * Included one more example program 'hlc' that show
     how to downcase all tags in an HTML file.

3.26     2002-03-17
  * Avoid core dump in some cases where the callback croaks.
     The perl_call_method and perl_call_sv needs G_EVAL flag
     to be safe.
  * New parser attributes; 'attr_encoded' and 'case_sensitive'.
     Contributed by Guy Albertelli II <>.
  * HTML::Entities
         - don't encode \r by default as suggested by Sean M. Burke.
  * HTML::HeadParser
         - ignore empty http-equiv
         - allow multiple <link> elements.  Patch by
           Timur I. Bakeyev <>
  * Avoid warnings from bleadperl on the uentities test.

3.25     2001-05-11
  * Minor tweaks for build failures on perl5.004_04, perl-5.6.0,
     and for macro clash under Windows.
  * Improved parsing of <plaintext>...  :-)

3.24     2001-05-09
  * $p->parse(CODE)
  * New events: start_document, end_document
  * New argspecs: skipped_text, offset_end
  * The offset/line/column counters was not properly reset
     after eof.

3.23     2001-05-01
  * If the $p->ignore_elements filter did not work as it should if
     handlers for start/end events was not registered.

3.22     2001-04-17
  * The <textarea> element is now parsed in literal mode, i.e. no other tags
     recognized until the </textarea> tag is seen.  Unlike other literal elements,
     the text content is not 'cdata'.
  * The XML &apos; entity is decoded.  It apos-char itself is still encoded as
     &#39; as &apos; is not really an HTML tag, and not recognized by many HTML

3.21     2001-04-10
  * Fix a memory leak which occurred when using filter methods.
  * Avoid a few compiler warnings (DEC C):
        - Trailing comma found in enumerator list
        - "unsigned char" is not compatible with "const char".
  * Doc update.

3.20     2001-04-02
  * Some minor documentation updates.

3.19_94     2001-03-30
  * Implemented 'tag', 'line', 'column' argspecs.
  * HTML::PullParser doc update.
     eg/hform is an example of HTML::PullParser usage.

3.19_93     2001-03-27
  * Shorten 'report_only_tags' to 'report_tags'.
     I think it reads better.
  * Bleadperl portability fixes.

3.19_92     2001-03-25
  * HTML::HeadParser made more efficient by using 'ignore_elements'.
  * HTML::LinkExtor made more efficient by using 'report_only_tags'.
  * HTML::TokeParser generalized into HTML::PullParser.  HTML::PullParser
     only support the get_token/unget_token interface of HTML::TokeParser,
     but is more flexible because the information that make up an token
     is customisable.  HTML::TokeParser is made into an HTML::PullParser

3.19_91     2001-03-19
  * Array references can be passed to the filter methods.  Makes it easier
     to use them as constructor options.
  * Example programs updated to use filters.
  * Reset ignored_element state on EOF.
  * Documentation updates.
  * The netscape_buggy_comment() method now generates mandatory warning
     about its deprecation.

3.19_90     2001-03-13
  * This is an developer only release.  It contains some new
     experimental features.  The interface to these might still change.
  * Implemented filters to reduce the numbers of callbacks generated:
        - $p->ignore_tags()
        - $p->report_only_tags()
        - $p->ignore_elements()
  * New @attr argspec.  Less overhead than 'attr' and allow
     compatibility with XML::Parser style start events.
  * The whole argspec can be wrapped up in @{...} to signal
     flattening.  Only makes a difference when the target is an

3.19     2001-03-09
  * Avoid the entity2char global.  That should make the module
     more thread safe.   Patch by Gurusamy Sarathy <>.

3.18     2001-02-24
  * There was a C++ style comment left in util.c.  Strict C
     compilers do not like that kind of stuff.

3.17     2001-02-23
  * The 3.16 release broke MULTIPLICITY builds.  Fixed.

3.16     2001-02-22
  * The unbroken_text option now works across ignored tags.
  * Fix casting of pointers on some 64 bit platforms.
  * Fix decoding of Unicode entities.  Only optionally available for
     perl-5.7.0 or better.
  * Expose internal decode_entities() function at the Perl level.
  * Reindented some code.

3.15     2000-12-26
  * HTML::TokeParser's get_tag() method now takes multiple
     tags to match.  Hopefully the documentation is also a bit clearer.
  * #define PERL_NO_GET_CONTEXT: Should speed up things for thread
     enabled versions of perl.
  * Quote some more entities that also happens to be perl keywords.
     This avoids warnings on perl-5.004.
  * Unicode entities only triggered for perl-5.7.0 or higher.

3.14     2000-12-03
  * If a handler triggered by flushing text at eof called the
     eof method then infinite recursion occurred.  Fixed.
     Bug discovered by Jonathan Stowe <>.
  * Allow <!doctype ...> to be parsed as declaration.

3.13     2000-09-17
  * Experimental support for decoding of Unicode entities.

3.12     2000-09-14
  * Some tweaks to get it to compile with "Optimierender Microsoft (R)
     32-Bit C/C++-Compiler, Version 12.00.8168, fuer x86."
     Patch by Matthias Waldorf <>.
  * HTML::Entities documentation spelling patch by
     David Dyck <>.

3.11     2000-08-22
  * HTML::LinkExtor and eg/hrefsub now obtain %linkElements from
     the HTML::Tagset module.

3.10     2000-06-29
  * Avoid core dump when stack gets relocated as the result of
    text handler invocation while $p->unbroken_text is enabled.
    Needed to refresh the stack pointer.

3.09     2000-06-28
  * Avoid core dump if somebody clobbers the aliased $self argument of
    a handler.
  * HTML::TokeParser documentation update suggested by
    Paul Makepeace <>.

3.08     2000-05-23
  * Fix core dump for large start tags.
    Bug spotted by Alexander Fraser <>
  * Added yet another example program: eg/hanchors
  * Typo fix by Jamie McCarthy <>

3.07     2000-03-20
  * Fix perl5.004 builds (was broken in 3.06)
  * Declaration parsing mode now only triggers for <!DOCTYPE ...> and
     <!ENTITY ...>.  Based on patch by la mouton <>.

3.06     2000-03-06
  * Multi-threading/MULTIPLICITY compilation fix.
     Both Doug MacEachern <> and
     Matthias Urlichs <> provided a patch.
  * Avoid some "statement not reached" warnings from picky
  * Remove final commas in enums as ANSI C does not allow
     them and some compilers actually care.
     Patch by James Walden <>
  * Added eg/htextsub example program.

3.05     2000-01-22
  * Implemented $p->unbroken_text option
  * Don't parse content of certain HTML elements as CDATA when
     xml_mode is enabled.
  * Offset was reported with wrong sign for text at end of chunk.

3.04     2000-01-15
  * Backed out 3.03-patch that checked for legal handler and attribute
    names in the HTML::Parser constructor.
  * Documentation typo fixed by Michael.

3.03     2000-01-14
  * We did not get out of comment mode for comments ending with an
    odd number of "-" before ">".  Patch by la mouton <>
  * Documentation patch by Michael.

3.02     1999-12-21
  * Hide ~-magic IV-pointer to 'struct p_state' behind a reference.
    This allow copying of the internal _hparser_xs_state element, and
    will make HTML-Tree-0.61 work again.
  * Introduced $p->init() which might be useful for subclasses that
    only want the initialization part of the constructor.
  * Filled out DIAGNOSTICS section of the HTML::Parser POD.

3.01     1999-12-19
  * Rely on ~-magic instead of a DESTROY method to deallocate
    the internal 'struct p_state'.  This avoid memory leaks
    when people simply wipe of the content of the object hash.
  * One of the assertion in hparser.c had opposite logic.  This made
    the parser fail when compiled with a -DDEBUGGING perl.
  * Don't assume any specific order of hash keys in the t/cases.t.
    This test failed with some newer development releases of perl.

3.00     1999-12-14
  * Documentation update (most of it from Michael)
  * Minor patch to eg/hstrip so that it use a "" handler
    instead of &ignore.
  * Test suite patches from Michael

2.99_96     1999-12-13
  * Patches from Michael:
    - A handler of "" means that the event will be ignored.
      More efficient than using 'sub {}' as handler.
    - Don't use a perl hash for looking up argspec keywords.
    - Documentation tweaks.

2.99_95     1999-12-09
  * (this is a 3.00 candidate)
  * Fixed core dump when "<" was followed by an 8-bit character.
    Spotted and test case provided by Doug MacEachern.  Doug had
    been running HTML-Parser-XS through more that 1 million urls that
    had been downloaded via LWP.
  * Handlers can now invoke $p->eof to request the parsing to terminate.
    HTML::HeadParser has been simplified by taking advantage of this.
    Also added a title-extraction example that uses this.
  * Michael once again fixed my bad English in the HTML::Parser
  * netscape_buggy_comment will carp instead of warn
  * updated TODO/README
  * Documented that HTML::Filter is depreciated.
  * Made backslash reserved in literal argspec strings.
  * Added several new test scripts.

2.99_94     1999-12-08
  * (should almost be a 3.00 candidate)
  * Renamed 'cdata_flag' as 'is_cdata'.
  * Dropped support for wrapping callback handler and argspec
    in an array and passing a reference to $p->handler.  It
    created ambiguities when you want to pass a array as
    handler destination and not update argspec.  The wrapping
    for constructor arguments are unchanged.
  * Reworked the documentation after updates from Michael.
  * Simplified internal check_handler().  It should probably simply
    be inlined in handler() again.
  * Added argspec 'length' and 'undef'
  * Fix statement-less label.  Fix suggested by Matthew Langford
  * Added two more example programs: eg/hstrip and eg/htext.
  * Various minor patches from Michael.

2.99_93     1999-12-07
  * Documentation update
  * $p->bool_attr_value renamed as $p->boolean_attribute_value
  * Internal renaming: attrspec --> argspec
  * Introduced internal 'enum argcode' in hparser.c
  * Added eg/hrefsub

2.99_92     1999-12-05
  * More documentation patches from Michael
  * Renamed 'token1' as 'token0' as suggested by Michael
  * For artificial end tags we now report 'tokens', but not 'tokenpos'.
  * Boolean attribute values show up as (0, 0) in 'tokenpos' now.
  * If $p->bool_attr_value is set it will influence 'tokens'
  * Fix for core dump when parsing <a "> when $p->strict_names(0).
    Based on fix by Michael.
  * Will av_extend() the tokens/tokenspos arrays.
  * New test suite script by Michael: t/attrspec.t

2.99_91     1999-12-04
  * Implemented attrspec 'offset'
  * Documentation patch from Michael
  * Some more cleanup/updated TODO

2.99_90     1999-12-03
  * (first beta for 3.00)
  * Using "realloc" as a parameter name in grow_tokens created
    problems for some people.  Fix by Paul Schinder <>
  * Patch by Michael that makes array handler destinations really work.
  * Patch by Michael that make HTML::TokeParser use this.  This gave a
    a speedup of about 80%.
  * Patch by Michael that makes t/cases into a real test.
  * Small HTML::Parser documentation patch by Michael.
  * Renamed attrspec 'origtext' to 'text' and 'decoded_text' to 'dtext'
  * Split up Parser.xs.  Moved stuff into hparser.c and util.c
  * Dropped html_ prefix from internal parser functions.
  * Renamed internal function html_handle() as report_event().

2.99_17     1999-12-02
  * HTML::Parser documentation patch from Michael.
  * Fix memory leaks in html_handler()
  * Patch that makes an array legal as handler destination.
    Also from Michael.
  * The end of marked sections does not eat successive newline
    any more.
  * The artificial end event for empty tag in xml_mode did not
    report an empty origtext.
  * New constructor option: 'api_version'

2.99_16     1999-12-01
  * Support "event" in argspec.  It expands to the name of the
    handler (minus "default").
  * Fix core dump for large start tags.  The tokens_grow() routine
    needed an adjustment.  Added test for this; t/largstags.t.

2.99_15     1999-11-30
  * Major restructuring/simplification of callback interface based on
    initial work by Michael.  The main news is that you now need to
    tell what arguments you want to be provided to your callbacks.
  * The following parser options has been eliminated:

2.99_14     1999-11-26
  * Documentation update by Michael A. Chase.
  * Fix for declaration parsing by Michael A. Chase.
  * Workaround for perl5.004_05 bug. Can't return &PL_sv_undef.

2.99_13     1999-11-22
  * New POD based on initial work by Michael A. Chase.
    All new features should now be described.
  * $p->callback(start => undef) will not reset the callback.
  * $p->xml_mode() did not parse attributes correct because
    HCTYPE_NOT_SPACE_EQ_SLASH_GT flag was never set.
  * A few more tests.

2.99_12     1999-11-18
  * Implemented $p->attr_pos attribute.  This causes attr positions
    within $origtext of the start tag to be reported instead of the
    attribute values.  The positions are reported as 4 numbers; end of
    previous attr, start of this attr, start of attr value, and end of
    attr.  This should make substr() manipulations of $origtext easy.
  * Implemented $p->unbroken_text attribute.  This makes sure that
    text segments are never broken and given back as separate text
    callbacks.  It delays text callbacks until some other markup
    has been recognized.
  * More English corrections by Michael A. Chase.
  * HTML::LinkExtor now recognizes even more URI attributes as
    suggested by Sean M. Burke <>
  * Completed marked sections support.  It is also now a compile
    time decision if you want this supported or not.  The only
    drawback of enabling it should be a possible parsing speed
    reduction.  I have not measured this yet.
  * The keys for callbacks initialized in the constructor are now
    suffixed with "_cb".
  * Renamed $p->pass_cbdata to $p->pass_self.
  * Added magic number to the p_state struct.

2.99_11     1999-11-17
  * Don't leak $@ modifications from HTML::Parser constructor.
  * Included HTML::Parser POD.
  * Marked sections almost work.  CDATA and RCDATA should work.
  * For tags that take us into literal_mode; <script>, <style>,
    <xmp>, we did not recognize the end tag unless it was written
    in all lower case.

2.99_10     1999-11-16
  * The mkhctype and mkpfunc scripts were using \z inside RE.  This
    did not work for perl5.004.  Replaced them with plain old
    dollar signs.

2.99_09     1999-11-15
  * Grammar fixes by Michael A. Chase <>
  * Some more test suite patches for Win32 by Michael A. Chase
  * Implemented $p->strict_names attribute.  By default we now
    allow almost anything in tag and attribute names.  This is much
    closer to the behaviour of some popular browsers.  This allows us
    to parse broken tags like this example from the LWP mailing list:
  * Introduced some tables in "hctype.h" and "pfunc.h".  These
    are built by the corresponding "mk..." script.

2.99_08     1999-11-10
  * Make Parser.xs compile on perl5.004_05 too.
  * New callback called 'default'.  This will be called for any
    document text no other callback shows an interest in.
  * Patch by Michael A. Chase <> that should
    help clean up files for the test suite on Win32.
  * Can now set up various attributes with key/value pairs passed to
    the constructor.
  * $p->parse_file() will open the file in binmode()
  * Pass complete processing instruction tag as second argument
    to process callback.
  * New boolean attribute v2_compat.  This influences how attributes
    are reported for start tags.
  * HTML::Filter now filters process instructions too.
  * Faster HTML::LinkExtor by taking advantage of the new
    callback interface.  The module now also uses (instead
    of the old URI::URL) to absolutize URIs.
  * Faster HTML::TokeParser by taking advantage of new
    accum interface.

2.99_07     1999-11-09
  * Entities in attribute values are now always expanded.
  * If you set the $p->decode_text_entities to a true value, then
    you don't have to decode the text yourself.
  * In xml_mode we don't report empty element tags as a start tag
    with an extra parameter any more.  Instead we generate an artificial
    end tag.
  * 'xml_mode' now implies 'keep_case'.
  * The parser now keeps its own copy of the bool_attr_value value.
  * Avoid memory leak for text callbacks
  * Avoid using ERROR as a goto label.
  * Introduced common internal accessor function for all boolean parser
  * Tweaks to make Parser.xs compile under perl5.004.

2.99_06     1999-11-08
  * Internal fast decode_entities().   By using it we are able to make
    the HTML::Entities::decode function 6 times faster than the old one
    implemented in pure Perl.
  * $p->bool_attr_value() can be set to influence the value that
    boolean attributes will be assigned.  The default is to assign
    a value identical to the attribute name.
  * Process instructions are reported as "PI" in @accum
  * $p->xml_mode(1) modifies how processing instructions are terminated
    and allows "/>" at the end of start tags.
  * Turn off optimizations when compiling with gcc on Solaris.  Avoids
    what we believe to be a compiler bug.  Should probably figure out
    which versions of gcc have this bug.

2.99_05     1999-11-05
  * The previous release did not even compile.  I forgot to try 'make test'
    before uploading.

2.99_04     1999-11-05
  * Generalized <XMP>-support to cover all literal parsing.  Currently
    activated for <script>, <style>, <xmp> and <plaintext>.

2.99_03     1999-11-05
  * <XMP>-support.
  * Allow ":" in tag and attribute names
  * Include rest of the HTML::* files from the old HTML::Parser
    package.  This should make testing easier.

2.99_02     1999-11-04
  * Implemented keep_case() option.  If this attribute is true, then
    we don't lowercase tag and attribute names.
  * Implemented accum() that takes an array reference.  Tokens are
    pushed onto this array instead of sent to callbacks.
  * Implemented strict_comment().

2.99_01     1999-11-03
  * Baseline of XS implementation

2.25     1999-11-05
  * Allow ":" in attribute names as a workaround for Microsoft Excel
    2000 which generates such files.
  * Make deprecate warning if netscape_buggy_comment() method is
    used.  The method is used in strict_comment().
  * Avoid duplication of parse_file() method in HTML::HeadParser.

2.24     1999-10-29
  * $p->parse_file() will not close a handle passed to it any more.
    If passed a filename that can't be opened it will return undef
    instead of raising an exception, and strings like "*STDIN" are not
    treated as globs any more.
  * HTML::LinkExtor knows about background attribute of <tables>.
    Patch by Clinton Wong <>
  * HTML::TokeParser will parse large inline strings much faster now.
    The string holding the document must not be changed during parsing.

2.23     1999-06-09
  * Documentation updates.

2.22     1998-12-18
  * Protect HTML::HeadParser from evil $SIG{__DIE__} hooks.

2.21     1998-11-13
  * HTML::TokeParser can now parse strings directly and does the
    right thing if you pass it a GLOB.  Based on patch by
    Sami Itkonen <>.
  * HTML::Parser now allows space before and after "--" in Netscape
    comments.  Patch by Peter Orbaek <>.

2.20     1998-07-08
  * Added HTML::TokeParser.  Check it out!

2.19     1998-07-07
  * Don't end a text chunk with space when we try to avoid breaking up

2.18     1998-06-22
  * HTML::HeadParser->parse_file will now stop parsing when the
    <body> starts as it should.
  * HTML::LinkExtor more easily subclassable by introducing the
    $self->_found_link method.

2.17     1998-04-28
  * Never split words (a sequence of non-space) between two invocations
    of $self->text.  This is just a simplification of the code that tried
    not to break entities.
  * HTML::Parser->parse_file now use smaller chunks as already
    suggested by the HTML::Parser documentation.

2.16     1998-04-02
  * The HTML::Parser could some times break hex entities (like &#xFFFF;)
    in the middle.
  * Removed remaining forced dependencies on libwww-perl modules.  It
    means that all tests should now pass, even if libwww-perl was not
    installed previously.
  * More tests.

2.14     1998-04-01
  * HTML::* modules unbundled from libwww-perl-5.22