REXML Change Log

Tagged for release 3.1.7

• Issue 14; 3.1.3: DTD events are now all being passed by StreamParser
• Issue #9; 3.1.3: The SAX parser was not denormalizing entity references in incoming text.
• Applied Curt Sampson's optimization improvements
• Fixed a buffering bug in Source.rb that affected the SAX parser
• Fixed a CDATA pretty-printing bug. (#39)

Tagged for release 3.1.7

r1287@bean: ser | 2007-07-24 20:12:25 -0400
Applied patch from Jeff Barczewski. Note that this changes what the values of
the name and IDs are from the previous behavior -- the values no longer include
the quotes. This is the correct behavior, so I'm leaving it in, but it is not
backwards compatible. Also fixes the serializer so that it outputs the doctype
in a correct format (needed as a result of this change).

Addresses ticket:92.

r1285@bean: ser | 2007-07-24 20:02:07 -0400
And THIS goes with the previous two patches. Dammit.

r1283@bean: ser | 2007-07-24 19:53:30 -0400
This goes with the previous commit.

r1281@bean: ser | 2007-07-24 11:08:48 -0400
Addresses ticket:85

This is a major rewrite of the XML formatting code. The XML writers have all
been extracted out of the classes and put into their own class containers.
This makes writing parsers easier, and cleaner.

There are three formatters, which correspond to the previous three XML writing
 modes:

REXML::Formatters::Default
Prints the XML document exactly as it was parsed
REXML::Formatters::Pretty
Pretty prints the XML document, destroying whitespace in the document
REXML::Formatters::Transitive
Pretty prints the XML document, preserving whitespace

All of the write() functions have been deprecated (some are still used, but
these will also go away) except the write() function on Document, which is left
for convenience. To pretty print an XML document the canonical way:

formatter = REXML::Formatters::Pretty.new( 5 ) # indent by 5 spaces
formatter.write( document, output )

r1279@bean: ser | 2007-06-09 23:19:02 -0400
Fixes ticket:89 -- encoding CP-1252 was broken. ISO-8859-15 had the same
 problem.

Also in this patch is a fix to merge.rb (unused, but it should at least
contain no errors), and a unit test for ticket:88.

This doesn't work yet.

* Adds :attribute_quoting to Element#context, allowing users to choose which

character to use to quote attribute values.
* Adds a REVISION attribute to the REXML object, for better version tracking
* Cross-merges some changes from the other branch

Documentation cleanup. Ticket:84

Fixes ticket:80

Adds a unit test for ticket:79

A test to confirm that ticket:77 is closed.

Cross-ported a fix for ticket:88 from CVS.

Addresses ticket:78

NOTE: that this also fixes what is technically another bug in REXML. REXML's
XPath parser used to allow exponential notation in numbers. The XPath spec
is specific about what a number is, and scientific notation is not included.
Therefore, this has been fixed.

Put in a test for ticket:75, which I've been unable to duplicate.

Fixes ticket:71

Addresses ticket:66

* Fixes ticket:61 (xpath_parser)
* Fixes ticket:63 (UTF-16; UNILE decoding was bad)
* Cleans up some tests, removing opportunities for test corruption
* Improves parsing error messages a little
* Adds the ability to override the encoding detection in Source construction
* Fixes an edge case in Functions::string, where document nodes weren't

correctly converted
* Fixes Functions::string() for Element and Document nodes
* Fixes some problems in entity handling

The changeset 1230:1231 was bad. The default behavior is *not* to use the
native REXML encodings by default, but rather to use ICONV by default.

Fixes a bug in multiple decodings

Two unrelated changes, because subversion is retarded and doesn't do
block-level commits:

1) Fixed a typo bug in previous change for ticket:15
2) Fixed namespaces handling in XPath and element.

***** Note that this is an API change!!! *****

Element.namespaces() now returns a hash of namespace mappings which are
relevant for that node.

Typo fix. Goes with 

This goes with ticket:68. Dammit.

Implements ticket:15, using gwrite's suggestion. This allows Element to be
subclassed.

Fixes ticket:68.

NOTE that this involves an API change! Entity declarations in the doctype now
generate events that carry two, not one, arguments.

Cross-patch from Ruby CVS; mostly Nabu edits.

* Adds a data file for a unit test that got missed
* Documentation updates

Fixed a really old crusty string from when importing stylesheets from 

ticket:21: Improved the error message for missing attribute quotes. There's
only so much we can do about this; if there's a missing or mismatched end
quote, how would we know?

Added a test to confirm ticket:76 has been resolved

Added unit test confirming ticket:57 has been fixed.

Changes some namespace logic. This is related to ticket:51.

If a namespace mapping is provided (during the call to match() or first()),
then this overrides the default nodal namespace mapping.

Added some unit tests for ticket:51. These will fail.

Another bit for ticket:52

Fix for ticket:52, spurious newlines added to comments.

Unit tests to confirm ticket:43 fixed

Unit tests added to confirm ticket:36

Resolves ticket:70. The root cause of this was the conversion of literals to
numbers and back to strings.

Unit test for feature request #16.

Implements feature request #16.

Adds a unit test specifically for ticket:60.

Fix for ticket:60, Functions::number not working correctly. The fix was provide
by Alex LeDonne, and included an excellent unit test. Alex's fix also fixes an
unfiled bug about number() not handling negative numbers.

Version bump

Adds a regression test for ticket #56, which was fixed
in an earlier, unidentified, commit.

Good patch by Aaron to fix ticket #53: REXML ignoring unbalanced tags
at the end of a document.

Various fixes to recent (unreleased) regressions

Fixes ticket #42

Adds a unit test for ticket #39. I don't know which changeset addressed this
issue.

Adds a test to confirm that ticket #12 has been fixed

Added the Jaxen test suite, contributed by Ryan Cox. Some rewriting was

necessary to merge it into the test framework, but most of the effort was 
contributed.

Formatting changes

Addresses ticket #10, fixing the StreamParser API for DTDs.

Re-enables a SAX test, and adds a new one.

Adds a test for ticket#58 and some documentation to clarify how XMLDecl
works, to avoid invalid bug reports.

Resolves ticket #49

Ticket #45
Now parses notation declarations in DTDs properly.

Ticket #31 - One unescape too many
This wasn't really a bug, per se... "value" always returns
a normalized string, and "value" is the method used to get
the text() of an element. However, entities have no meaning
in CDATA sections, so there's no justification for value
to be normalizing the content of CData objects. This behavior
has therefore been changed.

Renamed some unit tests to make them easier to identify and get at.

This test included a big XML file, which has been compressed.

Closes #26, allowing anything that walks like an IO
to be a source.

Applies a patch by David Tardon, which (incidentally) fixes 

Whitespace changes

Improved doctype handling

This undoes the patch for issue:48, and fixes it with cleaner whitespace
handling outside of the document context. The issue was that REXML was
stripping out and ignoring whitespace, which made it difficult to preserve
the structure of the document. This fixes that.

I'm not sure that this had anything to do with issue:48, but I was unable
to repeat the errors from issue:48, and it may be that, because of the
alternate encoding and ISO-9660-3 characters, the XML declaration was being
lost on repeated reads.

Merged encoding changes from CVS. Need to verify this,
as it may be causing problems.

Deals with carriage returns after XML 

This patch goes with 

Encodes some characters and removes a brokes link in the 

Fixes issue:46, adding methods for accessing notation DTD 

Fixes issue:48, repeated writes munging text 

* Moves parser.source.position() to parser.position()
* Improves the build script (less work for me to package a distribution)

* Resolves issue #34

* Changes Date, Version, and Copyright to upper case, to avoid conflicts with
the Date class. All of the other changes in the altered files are because
Subversion doesn't allow block-level commits, like it should.
* Minor, yet incomplete, documentation changes. Again, these are in this patch
because of Subversion's glaring lack of block-level commits.
* I forgot to add the unit tests that I promised in the previous patch for
issue #32.

* Merges a user-contributed patch for issue #40
* Adds a forgotten-to-commit unit test for issue #32

* Fixes ticket #32

* Fixes typo from previous commit

* Aliases #size to #length, as per zdennis's request.

Fixes ticket:30, XPath default namespace bug. The fix was provided
by Lucas Nussbaum.

Patch submitted by an anonymous doner to allow parsing of Tempfiles. I was

hoping that, by now, that whole Source thing would have been changed to use
duck typing and avoid this sort of issue... but in the meantime, the patch
has been applied.

Fixed issue44, Element#xpath

Fixes roundup issue 43: substring-after bug.
See: http://www.germane-software.com/cgi-bin/roundup/rexml/issue43

* More XPath ordering testing added
* Fixed the documentation WRT the raw mode of text nodes (ticket:4)

Cross ported fix from CVS

Version bump. File renaming.

Fixed a couple of bugs in the SAX2 parser, plus a bunch of other changes I don't remember.

Fixes ticket:3 (Issue38 in Roundup.)
However, this needs further testing.

require 

Fixes 

Wasn't including Text class.

Backed out the patch that changed the versions

* Changes by Matz, mostly of indent -> indent_level, to avoid
function/variable naming conflicts
* Tabs -> spaces (whitespace)

Indentation changes, and changed some exceptions to runtime
exceptions.

Fixed the inheritance bug in the pull parser that James Britt found.

Applied Nobu's changes to the Encoding infrastructure, which should fix

potential threading issues.

Added more tests, and the missing syncenumerator class.

Uncommented some code commented for testing

* Fixed preceding:: and following:: axis
* Fixed the ordering bug that Martin Fowler reported.

Merged the changes in from branches/xpath_V

Added Michael's sample code.

Added functions courtesy of Michael Neumann <mneumann@xxxx.de>.
Example code to follow.

Fixed a broken link. Thanks to Dick Davies for pointing it out.

* #29: XPath ordering bug fixed by Mark Williams. Incidentally, Mark supplied
a superlative bug report, including a full unit test. Then he went ahead and
fixed the bug. It doesn't get any better than this, folks.

* #26: Element#add_element(nil) now raises an error
* Changed XPath searches so that if a non-Hash is passed, an error is raised
* Fixed a spurrious undefined method error in encoding.

Some of the DTD events were not being passed through by the stream parser.

Setting bin apps to executable

bin/* programs should now be executable.

All declared internal entities, as well as numeric entities,
should now be denormalized. There was a related bug in that the SAX parser
was actually double-encoding entities; this is also fixed.

* Minor pretty printing fix.

The unit test for the previous patch. Forgot to commit it.

This bug was related to how REXML determines the encoding of a file, and
evinced itself by hanging on input when using the SAX parser.

* Previous fix, (include? -> includes?) was incorrect.
* Added another test for encoding
* Started AnyName support in RelaxNG
* Added Element#Attributes#to_a, so that it does something intelligent.
This was needed by XPath, for '@*'
* Fixed XPath so that @* works.
* Added xmlgrep to the bin/ directory. A little tool allowing you to grep
for XPaths in an XML document.

* Fixed bug #34, typo in xpath_parser.

* Minor big in missing includes for validation rules

* Removed debugging output from unit test
* Moved ">" in Element.inspect

* Redesigned and rewrote the RelaxNG code. It isn't elegant, but it works.
Particular problems encountered were interleave and ref. Interleave means I
can't use a clean FSM design, and ref means the dirty FSM design has to be
 modified
during validation. There's a lot of code that could be cleaned up in here.
However, I'm pretty sure that this design is reasonably fast and space
 efficient.
I'm not entirely convinced that it is correct; more tests are required.
* This version adds support for defines and refs.

PREPARES FOR DEF/REF. REMOVEABLE 

I forgot to add Kou's tests for the preceding-sibling patch.

* Applied Kou's preceding-sibling patch, which fixes the order of the axe results

* Added Kou's patch to normalize attribute values passed through the SAX2 and
Stream parsers.

* Added suppert for mixed

* Entirely rewrote the validation code; the finite state machine, while cool,
didn't survive the encounter with Interleave. It was getting sort of hacky,
too. The new mechanism is less elegant, but is basically still a FSM, and is
more flexible without having to add hacks to extend it. Large chunks of the
FSM may be reusable in other validation mechanisms.
* Added interleave support

Workin' in the coal mine, goin' down, down, down...

Version bump

* Added support for <value>

* Fixed a bunch of cases
* Added support for <group>

* Fixed text in choices. This change improves handling of all events that occur

without an end step (which is most of them).

Validation improvements:
* Fixed text
* Fixed attributes in choices

* XPath wasn't parsing ")" correctly.

Improved support for converting XPaths to strings.