grouping into element with condition on attribute of following sibling - xslt-2.0

Question

I have this xml-File:
<w>
<!-- many text nodes like below-->
<text attr1="3" width="100">This is a sentence.</text>
<text attr1="5" width="110">Another sentence, this time with a % in it.</text>
<text attr1="9" width="40">Some text.</text>
<text attr1="3" width="49">Other text.</text>
<text attr1="1" width="90">Again some text</text>
<!-- many text nodes like above-->
</w>
I want to collapse all following nodes of the element with the '%' in it which have a width-attribute lower than 50 into its text node and surround the whole group with <tag1>, so that it looks like this:
<w>
<!-- many text nodes like below-->
<text attr1="3" width="100">This is a sentence.</text>
<text attr1="5" width="110">Another sentence, this time with a
<tag1>% in it. Some text. Other text.</tag1>
</text>
<text attr1="1" width="90">Again some text</text>
<!-- many text nodes like above-->
</w>
I have tried it with this template:
<xsl:choose>
<xsl:when test="contains(.,'%') and following-group|#width <50">
<!-- I cannot choose current-group(), it generates a output with no tag1's
at all -->
<tag1><xsl:value-of-select="."/></tag1>
</xsl:when>
</xsl:choose>
But this only puts the line which contains the % in the <tag1>s. I dont understand why I cant select the whole group I match in the test. I am also aware that even if it matched the whole thing, the <tag1>s would include the whole line with the %, but this is secondary.

Answer 1

Here is a sample that tries to solve it, although I would need more input samples or clearer and more detailed description of the possible inputs:
<xsl:stylesheet
version="2.0"
xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
xmlns:xs="http://www.w3.org/2001/XMLSchema"
exclude-result-prefixes="xs">
<xsl:output method="xml" indent="yes"/>
<xsl:template match="#* | node()">
<xsl:copy>
<xsl:apply-templates select="#* , node()"/>
</xsl:copy>
</xsl:template>
<xsl:template match="w">
<xsl:copy>
<xsl:for-each-group select="*" group-starting-with="text[contains(., '%')]">
<xsl:choose>
<xsl:when test="not(self::text[contains(., '%')])">
<xsl:apply-templates select="current-group()"/>
</xsl:when>
<xsl:otherwise>
<xsl:variable name="head" select="."/>
<xsl:for-each-group select="current-group() except $head" group-adjacent="boolean(#width < 50)">
<xsl:choose>
<xsl:when test="current-grouping-key()">
<xsl:element name="{name($head)}" namespace="{namespace-uri($head)}">
<xsl:copy-of select="#*"/>
<xsl:analyze-string select="$head" regex="%.*$">
<xsl:matching-substring>
<tag1>
<xsl:value-of select="., current-group()/node()/string()"/>
</tag1>
</xsl:matching-substring>
<xsl:non-matching-substring>
<xsl:value-of select="."/>
</xsl:non-matching-substring>
</xsl:analyze-string>
</xsl:element>
</xsl:when>
<xsl:otherwise>
<xsl:apply-templates select="current-group()"/>
</xsl:otherwise>
</xsl:choose>
</xsl:for-each-group>
</xsl:otherwise>
</xsl:choose>
</xsl:for-each-group>
</xsl:copy>
</xsl:template>
</xsl:stylesheet>
With that stylesheet Saxon 9 transforms the input sample
<w>
<text>foo</text>
<text>bar</text>
<text attr1="3" width="100">This is a sentence.</text>
<text attr1="5" width="110">Another sentence, this time with a % in it.</text>
<text attr1="9" width="40">Some text.</text>
<text attr1="3" width="49">Other text.</text>
<text attr1="1" width="90">Again some text</text>
<text>foo bar</text>
<text>foo baz</text>
</w>
into the following result:
<w>
<text>foo</text>
<text>bar</text>
<text attr1="3" width="100">This is a sentence.</text>
<text attr1="9" width="40">Another sentence, this time with a <tag1>% in it. Some text. Other text.</tag1>
</text>
<text attr1="1" width="90">Again some text</text>
<text>foo bar</text>
<text>foo baz</text>
</w>