Question

[EDIT: Changed title to better conceptualize the question.]

The value of the attribute @xml:space can be either "default" or "preserve". XML specifies what the second means but leaves the first up to the application. (I think I have that correct.) So what if the application wants default to implement XSchema’s collapse? How could XSLT 1.0 actually do this?

I think the built-in template for processing text, that is,

<xsl:template match="text()">
   <xsl:value-of select="."/>
</xsl:template>

would need to be replaced with something like this pseudo-code:

<xsl:choose>
   <xsl:when test="../@xml:space='preserve'"
     <xsl:value-of select="."/>
   </xsl:when>
   <xsl:otherwise>

      if position(.)=1 then output LTRIM(value-of(.))
      if position(.)=last() then output RTRIM(value-of(.))
      if position(.)= 1 and last()=1 then output normalize-space(.)

   </xsl:otherwise>
</xsl:choose>

This input then:

<persName> The man is 
   <forename>Edward</forename>

   <forename>George</forename>
   <surname type="linked">Bulwer-Lytton</surname>, <roleName>Baron Lytton of
   <placeName>Knebworth</placeName>
   </roleName>
</persName>

would get rendered correctly as The man is Edward George Bulwer-Lytton, Baron Lytton of Knebworth with the space before The man and after Knebworth trimmed and the spaces between Edward and George collapsed. (The example is from TEI.)

[EDIT: I removed an incorrect and misleading paragraph here.]

The XSLT 1.0 to implement that pseudo-code would need to be executed for every text node. Wouldn’t that be ugly and slow? [EDIT: Or maybe not. I simplified the pseudo code. Are there fast trim routines? Is the choose really that slow?]

Bottom line: How does one implement XSchema’s collapse in XSLT 1.0 (with only browser-embedded extensions)?

I hope I’m saying all that correctly. And I hope the code is simple. I haven’t yet seen how it can be. [EDIT: Changed xs:collapse to XSchema’s collapse.]

Answer 1

I checked at xml-dev, and it turns out I was correct about the meaning and intended use of @xml:space.

Here is code to normalize whitespace in mixed-content elements (that’s a better way to say what I wanted to do):

<xsl:template priority=".7" match="text()[position()=1 and not((ancestor::node()/@xml:space)[position()=last()]='preserve')]">
    <xsl:value-of select="normalize-space()"/>
    <xsl:if test="normalize-space(substring(., string-length(.))) = ''">
        <xsl:text> </xsl:text>
    </xsl:if>
</xsl:template>
<xsl:template priority=".7" match="text()[position()=last() and not((ancestor::node()/@xml:space)[position()=last()]='preserve')]">
    <xsl:if test="normalize-space(substring(., 1, 1)) = ''">
        <xsl:text> </xsl:text>
    </xsl:if>
    <xsl:value-of select="normalize-space()"/>
</xsl:template>
<xsl:template priority=".8" match="text()[position()=1 and position()=last() and not((ancestor::node()/@xml:space)[position()=last()]='preserve')]" >
    <xsl:value-of select="normalize-space(.)"/>
</xsl:template>

Filtering on @xml:space allows preserve to override. The test= is just a way to test for whitespace. The priorities resolve the conflict caused when a node is the only text node in an element, and thus both the first and the last.