A preserveOrder= of "false" does nothing

[sydb]

In an ODD, <sequence preserveOrder="false"> produces the same RELAX NG as <sequence> (or <sequence preserveOrder="true">).

For testing, try this tiny ODD that tries to constraint <choice> to contain 1 <expan> and 1 <abbr> in either order. Then try it changing "false" to "true". (Note that, to appease GitHub, I have added a “.txt” extension to the filename.)

[sydb]

N.B. TEI issue #1970.

[martindholmes]

Pinging this ticket -- surely we need to get this fixed? It's currently not possible (as far as I can see) to create a content model which may contain zero or one of each of a list of elements in any order.

[joeytakeda]

I've recently been bitten by this bug and confirm it's still happening.

I've done a bit of digging in the Stylesheets and it looks like the root of the problem (or, at least, the initial problem) is in odd2odd.xsl. Running this command against the dev branch of the Stylesheets (where preserveOrder.odd is the ODD @sydb provided):

saxon -s:preserveOrder.odd -xsl:../Stylesheets/odds/odd2odd.xsl -o:preserveOrder_expanded.odd

Shows a <sequence> for content that has lost its @preserveOrder entirely

          <content>
            <sequence minOccurs="1" maxOccurs="1">
              <elementRef key="abbr"/>
              <elementRef key="expan"/>
            </sequence>
          </content>

I think the problem is in the odd2odd-simplifyODD template and, in particular, this condition:

Stylesheets/odds/odd2odd.xsl

Lines 1489 to 1493 in 9600186

	<xsl:when test="$entCount gt 0 or $stripped='true'">
	<xsl:element namespace="http://www.tei-c.org/ns/1.0" name="{$element}">
	<xsl:attribute name="minOccurs" select="$min"/>
	<xsl:attribute name="maxOccurs" select="if ($max eq -1) then 'unbounded' else $max"/>
	<xsl:copy-of select="@*|*|text()|processing-instruction()"/>

Only the @minOccurs and @maxOccurs is retained from the context element and all other attributes are ignored; just did a quick test by adding @xml:id to the <sequence> and it's also absent from preserveOrder_expanded.odd.

I'm happy to work on this in a fork and file a PR, if that would be helpful!

[joeytakeda]

The problem was indeed in odd2odd.xsl as far as I can tell. teiodds.xsl (imported byodd2relax.xsl) already has handling for preserveOrder:

Stylesheets/odds/teiodds.xsl

Lines 2078 to 2104 in 9600186

	<xsl:template match="tei:sequence" mode="#default tangle">
	<!-- "owe" = occurence wrapper element -->
	<xsl:variable name="owe" select="tei:generateIndicators(@minOccurs,@maxOccurs)"/>
	<!-- sequences of <dataRef> need use <list>, not <group> -->
	<xsl:variable name="group_or_list" select="if ( *[ not( self::tei:dataRef ) ] ) then 'group' else 'list'"/>
	<xsl:choose>
	<xsl:when test="@preserveOrder eq 'false' and string-length($owe) eq 0">
	<xsl:element name="{$group_or_list}" namespace="http://relaxng.org/ns/structure/1.0">
	<rng:interleave>
	<xsl:apply-templates mode="tangle"/>
	</rng:interleave>
	</xsl:element>
	</xsl:when>
	<xsl:when test="string-length($owe) eq 0">
	<xsl:element name="{$group_or_list}" namespace="http://relaxng.org/ns/structure/1.0">
	<xsl:apply-templates mode="tangle"/>
	</xsl:element>
	</xsl:when>
	<xsl:otherwise>
	<xsl:element name="{$owe}" namespace="http://relaxng.org/ns/structure/1.0">
	<xsl:element name="{$group_or_list}" namespace="http://relaxng.org/ns/structure/1.0">
	<xsl:apply-templates mode="tangle"/>
	</xsl:element>
	</xsl:element>
	</xsl:otherwise>
	</xsl:choose>
	</xsl:template>

So copying over the attributes in odd2odd.xsl seems to fix the issue, preserving @preserveOrder so that it's recognized by the time the ODD hits the above template.

I've forked the repo and made the change to odd2odd.xsl (will do a PR once I post this comment) as well as added some tests to Test2. I wasn't sure if that was the proper way to add a test--I couldn't find any documentation on how to add tests to Test, but the documentation for Test2 was clear, so I just went ahead and did that; let me know if that's actually helpful and, if not, happy to remove the changes to Test2.

[sydb]

Merged in #507, and no errors on Test or Test2. Yay!
Thanks @joeytakeda!