xsl:analyze-string to work on latex file

pietromaria.liuzzo · November 9th, 2013, 08:00 AM

Dear all, I am again trying to use xslt for thing which I reckon it is not designed for: it works very well so often that I insist in doing that...

I have now a latex file which has thousands of entries like this

Code:

\index[p]{pino!1!115.3}

which need to get (According to kinds) into this format

Code:

\index[p]{[email protected]}

So I have used the following xsl

Code:

<xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
    version="2.0">
    <xsl:template match="/">
        <xsl:variable name="hdt">
        <xsl:call-template name="hdt"/>
        </xsl:variable>

        <xsl:for-each select="$hdt">
    <xsl:call-template name="plut"/>
</xsl:for-each>

    </xsl:template>
    
    
    
    <xsl:template name="hdt">
        <xsl:analyze-string select="." regex="([A-Za-z0-9]+)*(!)(\d)(!)((\d+)(\.)*(\d*)(\-*)(\d*)((\.)(\d*))*)(\|)*([A-Za-z0-9]+)*(\}})">
    <xsl:matching-substring>
        <xsl:value-of select="regex-group(1)"/>
        <xsl:value-of select="regex-group(2)"/>
        <xsl:value-of select="regex-group(3)"/>
        <xsl:value-of select="regex-group(4)"/>
            <xsl:value-of select="format-number(number(regex-group(6)), '0000')"/>
        <xsl:choose>
            <xsl:when test="regex-group(8)">
            <xsl:value-of select="format-number(number(regex-group(8)), '0000')"/>
        </xsl:when>
        <xsl:otherwise>
<xsl:text>0000</xsl:text>
        </xsl:otherwise></xsl:choose>
        <xsl:text>@</xsl:text>
        <xsl:value-of select="regex-group(5)"/>
        <xsl:value-of select="regex-group(14)"/>
        <xsl:value-of select="regex-group(15)"/>
        <xsl:value-of select="regex-group(16)"/>
    </xsl:matching-substring>
    <xsl:non-matching-substring>
        <xsl:value-of select="."/>
    </xsl:non-matching-substring>
</xsl:analyze-string>
    </xsl:template>
    <xsl:template name="plut">
        <xsl:analyze-string select="." regex="([A-Za-z0-9]+)*(!)(\\*)([A-Za-z0-9]+)*(\{{)*([A-Za-z0-9]+)*(\}})*((!)((\d+)(\.)*(\d*)(\-*)(\d*)((\.)(\d*))*)(\|)*([A-Za-z0-9]+)*)*(\}})">
            <xsl:matching-substring>
                <xsl:value-of select="regex-group(1)"/>
                <xsl:value-of select="regex-group(2)"/>
                <xsl:value-of select="regex-group(3)"/>
                <xsl:value-of select="regex-group(4)"/>
                <xsl:value-of select="regex-group(5)"/>
                <xsl:value-of select="regex-group(6)"/>
                <xsl:value-of select="regex-group(7)"/>
                <xsl:value-of select="regex-group(9)"/>
                
                <xsl:value-of select="format-number(number(regex-group(10)), '0000')"/>
                <xsl:choose>
                    <xsl:when test="regex-group(12)">
                        <xsl:value-of select="format-number(number(regex-group(12)), '0000')"/>
                    </xsl:when>
                    <xsl:otherwise>
                        <xsl:text>0000</xsl:text>
                    </xsl:otherwise></xsl:choose>
                <xsl:text>@</xsl:text>
                <xsl:value-of select="regex-group(10)"/>
                <xsl:value-of select="regex-group(19)"/>
                <xsl:value-of select="regex-group(20)"/>
                <xsl:value-of select="regex-group(21)"/>
            </xsl:matching-substring>
            <xsl:non-matching-substring>
                <xsl:value-of select="."/>
            </xsl:non-matching-substring>
        </xsl:analyze-string>
    </xsl:template>
</xsl:stylesheet>

This works very well on a small file with only a selection of all things I want to match, which I have used to test. It does not work when I try to run it on a larger file. Please, do you have any suggestion on why this happens? Thank you very much
Pietro

mhkay · November 9th, 2013, 01:08 PM

It's not so much that XSLT is unsuited to this task, as that regular expressions are unsuited to it. I've no idea if latex is actually a regular language (=one that is cabable of being parsed using regular expressions), but even if it is, a regular expression that allows backtracking will often have performance that increases exponentially with the size of the string to be parsed. (And I mean expontentially in the technical sense of the term.)

I can't advise you in detail without knowing the syntax of the language you are trying to analyse, but I would have thought a conventional (e.g. recursive descent) parsing approach without unlimited lookahead or backtracking would be the right approach.

Looking very quickly at your actual regex, it starts

([A-Za-z0-9]+)*

which is surely equivalent to

[A-Za-z0-9]*

Nested repetitions like this vastly increase the number of possible ways in which a regular expression can match a string, and therefore worsen the performance of the matching algorithm.