I am trying to group using <xsl:for-each-group> and tokenize() but have am not achieving the desired results. Better to look at the code.

Source XML:
<worksheet name="users">
  <user index="1">
    <cell name="User Name">Suzy Smith</cell>
    <cell name="role">developer</cell>
  </user>
  <user index="2">
    <cell name="User Name">Tom Case</cell>
    <cell name="role">project_manager;team_leader</cell>
  </user>
  <user index="3">
    <cell name="User Name">Ian Lloyd</cell>
    <cell name="role">project_manager;team_leader</cell>
  </user>
  <user index="4">
    <cell name="User Name">John Gerrard</cell>
    <cell name="role">project_manager;developer</cell>
  </user>
</worksheet>

I want to generate the following code, where users are grouped according to their role. For each role (project_manager, developer or team_leader), the result will be a group of users.

Desired Result XML:

<list>
<group>developer</group>
<user>Suzy Smith</user>
<user>John Gerrard</user>
</list>

The problem I am having is this. If all cell[@name='role'] values are delimited lists, then it works fine. Or if there is one value, but it is followed with a ';', it will be included in the same group the tokenized values. But when the value is a single role, it is processed separately and not included in the 'group'.

So what I am getting is something like this:

<list>
<group>developer</group>
<user>Suzy Smith</user>
</list>

<list>
<group>developer</group>
<user>John Gerrard</user>
</list>

The simplified version of the XSLT is:

<xsl:for-each-group select="user" group-by="tokenize(cell[@name='role'],';')">

  <xsl:element name="list">

    <xsl:element name="group">
      <xsl:value-of select="current-grouping-key()" />
    </xsl:element>


    <xsl:for-each select="current-group()/cell[@name='User Name'] ">
      <xsl:element name="user">
        <xsl:value-of select="." />
      </xsl:element>
    </xsl:for-each>

  </xsl:element>

</xsl:for-each-group>

I am using saxon 8. If anyone has any idea, would greatly appreciated it.

Your code works for me. With this stylesheet, using Saxon 8.9.0.3

<?xml version="1.0" encoding="iso-8859-1"?>
<xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
                xmlns:xs="http://www.w3.org/2001/XMLSchema"
                version="2.0">

<xsl:template match="worksheet">
    <xsl:for-each-group select="user" group-by="tokenize(cell[@name='role'],';')">

      <xsl:element name="list">

        <xsl:element name="group">
          <xsl:value-of select="current-grouping-key()" />
        </xsl:element>

        <xsl:for-each select="current-group()/cell[@name='User Name'] ">
          <xsl:element name="user">
            <xsl:value-of select="." />
          </xsl:element>
        </xsl:for-each>
      </xsl:element>
    </xsl:for-each-group>
</xsl:template>
</xsl:stylesheet>

I get this output:

<?xml version="1.0" encoding="UTF-8"?>
<list>
   <group>developer</group>
   <user>Suzy Smith</user>
   <user>John Gerrard</user>
</list>
<list>
   <group>project_manager</group>
   <user>Tom Case</user>
   <user>Ian Lloyd</user>
   <user>John Gerrard</user>
</list>
<list>
   <group>team_leader</group>
   <user>Tom Case</user>
   <user>Ian Lloyd</user>
</list>

Michael Kay
http://www.saxonica.com/
Author, XSLT Programmer's Reference and XPath 2.0 Programmer's Reference

Sorry, I cleaned up the XML input, and I should not have. I think it is a white space issue. Is there any way to normalize-space when tokenizing?

<xsl:for-each-group select="row" group-by="tokenize(cell[@name='gdmsrole'],';')">

>Is there any way to normalize-space when tokenizing?

You can apply normalize-space() to the input before tokenizing but that will leave single spaces. If you simply want to ignore whitespace after the semicolon, then include it in the regex: tokenize($in, ';\s*')

Michael Kay
http://www.saxonica.com/
Author, XSLT Programmer's Reference and XPath 2.0 Programmer's Reference

Forgive me for not fully understanding how to control this just yet. The input varies, therefore, I can't include whitespace handling in the regex.

One node:

<cell name="role" index="3">
coordinators;project_manager
</cell>

Another node to match for group (coordinators)

<cell name="role" index="3">
coordinators
</cell>

If I were to normalize-space on the input, just where would that fit into the stylesheet?