XSLT 1.0:
<xsl:stylesheet
version="1.0"
xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
>
<!-- index Description elements by their text value -->
<xsl:key name="kDescription" match="Description" use="text()" />
<xsl:template match="/">
<output>
<!-- process all Description elements... -->
<xsl:apply-templates select="Problems/Problem/Description">
<!-- ...sorted by their own text value, ascending -->
<xsl:sort select="text()" />
<!-- pass in the File value that we want to filter for -->
<xsl:with-param name="file" select="'file1'" />
</xsl:apply-templates>
</output>
</xsl:template>
<xsl:template match="Description">
<xsl:param name="file" select="''" />
<!--
check if the current Description node is the first in its
respective group, that has a File value we care for
-->
<xsl:if test="
$file != ''
and
generate-id()
=
generate-id(key('kDescription', .)[../File = $file][1])
">
<!-- for the sake of simplicity, just make a copy here -->
<xsl:copy-of select="." />
</xsl:if>
</xsl:template>
</xsl:stylesheet>
With this input:
<Problems>
<Problem>
<File>file1</File>
<Description>desc1</Description>
</Problem>
<Problem>
<File>file1</File>
<Description>desc2</Description>
</Problem>
<Problem>
<File>file2</File>
<Description>desc3</Description>
</Problem>
<Problem>
<File>file2</File>
<Description>desc1</Description>
</Problem>
<Problem>
<File>file1</File>
<Description>desc2</Description>
</Problem>
</Problems>
I get:
<output>
<Description>desc1</Description>
<Description>desc2</Description>
</output>
A brief explanation of the part of the stylesheet that does the heavy lifting:
$file != ''
and
generate-id()
=
generate-id(key('kDescription', .)[../File = $file][1])
The first part is obvious - it's just there to make sure that a $file filter string was passed in.
The second part is plain Muenchian grouping with a little twist. It compares the IDs of two nodes, the current one (generate-id()) and one from the kDescription group, filtered by $file value.
kDescription indexes <Description> elements by their text value, which means that nodes with the same text but a different accompanying <File> will be returned by the call to key(). We need to filter them out.
If the current node is equal to the first node of the group that has the right <File> value, the test succeeds and something gets printed, otherwise nothing happens.