The thing is, XPathNodeIterator itself is stateful and does not behave
like IEnumerable, thus GetEnumerator() should not iterate nodes within
*itself*. Rather, it should have returned a clone of it.
After all we didn't have to damage performance to fix bug #3705, which is nice.
A valid way to do equivalent, from IRC:
[12:05] <eno> miguel: I believe you shouldn't add non-standard "ends-with" as the standard XPath implementation. Use extension framework instead.
[12:05] <eno> regarding https://github.com/mono/mono/commit/9b07497f9067ab4d3f4a7964f2868c7eafef1b5d
[13:39] <migueld> Seems like it is XPath 2.0
[13:46] <eno> Yes, XPath 2.0 contains that one. .NET implementation is for 1.0.
[13:46] <eno> This would help you to use XSLT extension http://msdn.microsoft.com/en-us/library/tf741884.aspx
[13:47] <eno> or this might be handy: http://msdn.microsoft.com/en-us/library/wxaw5z5e.aspx
[13:49] <eno> Add this under <stylesheet> or <transform> : <msxsl:script language="C#" implements-prefix="x" xmlns:msxsl="urn:schemas-microsoft-com:xslt">public bool ends_with (string s, string target) { return s.EndsWith (target); }</msxsl:script>
[13:49] <eno> and use "x:ends_with(foo, bar)" instead
Adding a test case for ancestor-or-self along the lines of the existing one for ancestor showed runaway, memory-eating behaviour. In changing CollectResults to remove the livelock, it seemed simplest to remove the extra yield of the root from MoveNextCore. I was unable to ascertain a good reason for producing this element in this special way compared to the AncestorIterator.
According to http://www.w3.org/TR/xpath#predicates, a predicate [n] should be equivalent to [position()=n] for a number expression n. However, we currently have a PredicateIterator intepretation of [n] that has been corrected for the issue outlined in the bug report, leaving the position() function behind with an inappropriate, document-order value for reverse axes. Note that [n] is not treated by parsing or transforming to the same form as [position()=n] before evaluation, valid as that would be.
The discussion in that same section of the recommendation about forward and reverse axes and further remarks in http://www.w3.org/TR/xpath#node-sets make it clear that the predicate [1] has the sense of "nearest" rather than "first in the document" which then, depending on the axis type, may be the same or reverse of the document order. It does not seem to me to be so clear whether the order given by a navigator for a location path expression should mirror this or always give the document order. The .Net implementation seems to always give the document order, as well as predicates having the correct reverse/forward senses, so I have added test cases that explore both.
* DefaultContext.cs : do not use sloppy XmlConvert to parse xpath
number. Fixed bug #556619.
* XslTransformTests.cs : added test for bug #556619 by
Tor Lillqvist.
svn path=/trunk/mcs/; revision=146613
* Expression.cs : to support external function that returns non-
double (such as int) as number, add conversion from non-double at
evaluation. This should fix bug #498389.
svn path=/trunk/mcs/; revision=141965
* XPathNavigator.cs : name test did not handle String.Empty cases
for name and ns (i.e. no filter). Fixed bug #515136.
* XPathNavigatorTests.cs : added test for bug #515136.
svn path=/trunk/mcs/; revision=138384
* Expression.cs: improve the ExpressionCache, it still does not
work, but it will at least work better with multithreaded apps (we
take locks) and we use WeakReferences correctly.
For WeakReferences it is important to not depend on IsAlive to
retrieve the target (this is a documented race). IsAlive is only
useful to determine if it has already been collected, but if true,
the Target can still be released.
This race condition is exposed by #505678 on x86-64 platforms (only).
Sadly, this fix is not good enough to re-enable the cache, on
x86-64 this with a single thread, it still crashes randomly
svn path=/trunk/mcs/; revision=135815
* Expression.cs : gave up another optimization for bug #443090.
* SelectNodesTests.cs : added 2nd test for bug #443490.
svn path=/trunk/mcs/; revision=129574
* XPathExpression.cs : eliminate use of ExpressionCache and lose
significant performance in some use for the sake of multithreaded
use. Fixed bug #477049.
svn path=/trunk/mcs/; revision=129545
* Iterator.cs : DescendantIterator.Clone() was not cloning current
node in new code. Fixed bug #456103.
* XPathNavigatorTests.cs : added test fo bug #456103.
svn path=/trunk/mcs/; revision=121080
* XPathNavigator.cs : made some 2.0-only methods available in 1.x
so that we don't have to hesitate to optimize xpath using them.
svn path=/trunk/mcs/; revision=113491
* XPathNavigator.cs : fixed InnerXml to not give up with root node.
Indent InnerXml. Fixed bug #376191.
* XPathNavigatorTests.cs : added test for bug #376191.
svn path=/trunk/mcs/; revision=99623
* XPathNavigator.cs : in some MoveTo*() methods, it should allow any
destination node when "type" is XPathNodeType.All. Based on the
patch by Sanghyeon Seo, fixed bug #365112.
* XPathNavigatorTests.cs : added test for bug #365112.
svn path=/trunk/mcs/; revision=96753
* DefaultContext.cs : now number formatting for "R" works fine, so
just use it (and remove conditional compilation for TARGET_JVM).
svn path=/trunk/mcs/; revision=95479
* Iterator.cs : be more strict to detect the end of
PredicateIterator. Fixed bug #349111.
* XslTransformTests.cs : added test for bug #349111.
svn path=/trunk/mcs/; revision=91405
* Expression.cs : In any comparison expr, treat RTF as equivalent to
string type (as specified in XSLT 1.0 section 11.1: "An operation
is permitted on a result tree fragment only if that operation would
be permitted on a string." ...). Fixed bug #349035.
* XslTransformTests.cs : added test for bug #349035.
svn path=/trunk/mcs/; revision=91402
* Expression.cs : text() should match whitespace and significant
whitespace nodes.
* XPathNavigatorTests.cs : added test for text() and whitespace/
significantwhitespace node.
svn path=/trunk/mcs/; revision=70536
* XPathNavigator.cs :
- ReadSubtree() should reject everything but Element or Root.
- InnerXml and OuterXml should not use ReadSubtree() when it is
not positioned on an Element or the Root. Thus, implement them
in their own ways. Fixed bug #79875.
* XPathNavigatorTests.cs : added test for bug #79875, and some
"escaped" text tests for InnerXml.
* XPathNavigatorReaderTests.cs : make sure that most of XPath nodes
are rejected to read subtree.
svn path=/trunk/mcs/; revision=67659
* XPathNavigator.cs : InnerXml should allow Text-only content. Fixed
bug #79874, when tied to XPathNavigatorReader fix.
* XPathNavigatorReader.cs : removed node type restriction on
initialization. Fixed bug #79874, when tied to XPathNavigator fix.
* XPathNavigatorTests.cs : added test for bug #79874.
svn path=/trunk/mcs/; revision=67653
* XPathNavigatorCommonTests.cs : added test for bug #78067.
* XPathNavigatorReaderTests.cs : ditto, for ReadSubtree() part.
* XPathNavigatorReader.cs : Read() did not compute Depth correctly.
It caused bug #78067.
* XPathNavigator.cs : get_OuterXml() returns indented output, without
XML declaration.
* XmlWriter.cs : WriteNode(XPathNavigator) : directly write element
and root nodes instead of dispatching to ReadSubtree().
svn path=/trunk/mcs/; revision=59286
* XslTransformProcessor.cs,
Compiler.cs : eliminate ExpressionStore. Now sorting is done
inside XslSortEvaluator.
* XslSortEvaluator.cs : new file. It handles current node in XSLT
context. Fixed bug #77781.
* Expression.cs : extracted XPathSortElement, XPathSorters
and XPathSorter from CompiledExpression, and split
XPathSorters.Sort() into some methods. Those changes are to make
them reusable in XslSortEvaluator.
* System.Xml.dll.sources : added XslSortEvaluator.cs.
* knownFailures.txt : removed two xsl:sort+current() tests.
svn path=/trunk/mcs/; revision=58015
* Iterator.cs : seems like MS XPathNodeIterator.MoveNext() itself does
not handle XsltContext.PreserveWhitespace() unlike ours. It results
in not a small performance improvement. (just removed it since it
didn't cause any errors in XSLT standalone tests.)
svn path=/trunk/mcs/; revision=57298