Question

I’m processing a source HTML file that holds tabular data in an unstructured way. Basically it’s a bunch of absolutely positioned divs. My goal is to rebuild some sort of structured XML data. So far, using XSLT 2.0 I was able to produce an XML looking like this:

<data>
    <line top="44">
         <item left="294">Some heading text</item>
    </line>
    <line top="47">
         <item left="718">A</item> <!-- this item is a section-start -->
         <item left="764">Section heading</item>
    </line>
    <line top="78">
        <item left="92">Data</item>
        <item left="144">Data</item>
        <item left="540">Data</item>
        <item left="588">Data</item>
    </line>
    <line top="101">
        <item left="61">B</item> <!-- this item is a section-start -->
        <item left="144">Section heading</item>
    </line>
    <line top="123">
        <item left="92">Data</item>
        <item left="144">Data</item>
    </line>
</data>

However, what I need to do next is group lines into sections. Each section starts with a line whose first item’s value consists of a single letter A – Z. My approach is to hold all the <line> elements in a $lines variable and then use xsl:for-each-group with group-starting-with attribute to identify the element starting a new section.

The respective XSLT fragment looks like this:

<xsl:for-each-group select="$lines/line" group-starting-with="...pattern here...">
    <section>
        <xsl:copy-of select="current-group()"/>
    </section>
</xsl:for-each-group>

The problem is I can’t figure out a working pattern to identify section starts. The best I could do was ensuring that //line/item[1]/text()[matches(., '^[A-Z]$')] works when used separately in an XPath evaluator. However, I can’t seem to derive a working version to be used with group-starting-with.

Update Hence the wanted result should look like this:

<data>
    <section> <!-- this section started automatically because of being at the beginning -->
        <line top="44">
             <item left="294">Some heading text</item>
        </line>
    </section>
    <section>
        <line top="47">
             <item left="718">A</item> <!-- this item is a section-start -->
             <item left="764">Section heading</item>
        </line>
        <line top="78">
            <item left="92">Data</item>
            <item left="144">Data</item>
            <item left="540">Data</item>
            <item left="588">Data</item>
        </line>
    </section>
    <section>
        <line top="101">
            <item left="61">B</item> <!-- this item is a section-start -->
            <item left="144">Section heading</item>
        </line>
        <line top="123">
            <item left="92">Data</item>
            <item left="144">Data</item>
        </line>
    </section>
</data>

Answer 1

The solution:

<xsl:for-each-group select="$lines/line" group-starting-with="line[matches(child::item[1], '^[A-Z]$')]">
    <section name="{current-group()[1]/item[1]}">
        <xsl:copy-of select="current-group()"/>
    </section>
</xsl:for-each-group>

The trick is really understanding that group-starting-with shall be a pattern not a condition.