Optimize <xsl:choose> with many <xsl:when>

Question

I have a series of medium-sized XML documents, which are mainly text with a few nodes representing macros to be expanded, e.g.:

<foo>Some text <macro>A1</macro> ... <macro>B2</macro> ...etc...</foo>

My goal is to replace each macro with the corresponding XML. Usually it's a single <img> tag with different attributes, but it could be some other HTML as well.

The stylesheet is generated programatically, and one way to do that would be to have a template per macro, e.g.

<xsl:template match="macro[.='A1']">
    <!-- content goes here -->
</xsl:template>
<xsl:template match="macro[.='A2']">
    <!-- other content goes here -->
</xsl:template>
<xsl:template match="macro[.='B2']">
    <!-- etc... -->
</xsl:template>

It works just fine, but it can have up to a hundred macros and it's not very performant (I'm using libxslt.) I've tried a couple of alternative such as:

<xsl:template match="macro">
    <xsl:choose>
        <xsl:when test=".='A1'">
            <!-- content goes here -->
        </xsl:when>
        <xsl:when test=".='A2'">
            <!-- other content goes here -->
        </xsl:when>
        <xsl:when test=".='B2'">
            <!-- etc... -->
        </xsl:when>
    </xsl:choose>
</xsl:template>

It's slightly more performant. I have tried adding another level of branching such as:

<xsl:template match="macro">
    <xsl:choose>
        <xsl:when test="substring(.,1,1) = 'A'">
            <xsl:choose>
                <xsl:when test=".='A1'">
                    <!-- content goes here -->
                </xsl:when>
                <xsl:when test=".='A2'">
                    <!-- other content goes here -->
                </xsl:when>
            </xsl:choose>
        </xsl:when>
        <xsl:when test=".='B2'">
            <!-- etc... -->
        </xsl:when>
    </xsl:choose>
</xsl:template>

It loads slightly slower (the XSL is bigger and a bit more complicated) but it executes slightly faster (each branch can eliminate several cases.)

Now I am wondering, is there a better way to do that? I have around 50-100 macros. Normally, the transformation is executed using libxslt but I can't use proprietary extensions from other XSLT engines.

Any input is welcome :)

Answer 1

If your macros are fixed XML, you could have a macros.xml document that is something like:

<?xml version="1.0"?>
<macros>
    <macro name="A1"><!-- content goes here --></macro>
    <macro name="A2"><!-- content goes here --></macro>
</macros>

You can then have:

<xsl:template match="macro">
    <xsl:variable name="name" select="text()"/>
    <xsl:apply-templates select="document(macros.xml)/macros/macro[@name=$name]" mode="copy"/>
</xsl:template>

<xsl:template match="*" mode="copy">
    <xsl:copy><xsl:copy-of select="*"/>
        <xsl:apply-templates mode="copy"/>
    </xsl:copy>
</xsl:template>

Does this improve the performance?

Answer 2

Try extracting all the macro processing into a separate template mode, and only run it for the contents of macro element. I.e.:

<xsl:template match="macro">
   <xsl:apply-templates mode="macro"/>
</xsl:template>

<xsl:temlate match="text()[. = 'A1']" mode="macro">
   ...
</xsl:template>

I suspect the slowdown in your case is because your rules get checked, one by one, for every node in input. This way, you get one check to see if it's a macro, and only if it is does it content get matched further.

Answer 3

This would be another way:

<xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform" exclude-result-prefixes="xsl">
    <xsl:variable name="dummy">
            <mac id="A1">success</mac>
            <mac id="A2">fail</mac>
            <mac id="B1">This <b>fail</b></mac>
            <mac id="B2">This <b>success</b></mac>
    </xsl:variable>
    <xsl:key name="macro" match="mac" use="@id"/>
    <xsl:template match="@*|node()">
        <xsl:copy>
            <xsl:apply-templates select="@*|node()"/>
        </xsl:copy>
    </xsl:template>
    <xsl:template match="macro">
        <xsl:variable name="me" select="."/>
        <xsl:for-each select="document('')">
            <xsl:copy-of select="key('macro',$me)/node()"/>
        </xsl:for-each>
    </xsl:template>
</xsl:stylesheet>

Note: This was the performance: XML parse 1,805ms, XSL parse 0,483ms, Transform 0,215ms