I encountered a problem today with the DocBook stylesheet's handling of
recursive XIncludes with the Xerces parser.
DocBook assumes that xml:base attributes are declared relative to one
another. Currently, it combines the path components of all xml:base
attributes recursively until it finds an absolute Uri or reaches the root.
However, at least the Xerces XInclude processor doesn't work like this. The
xml:base attributes are always relative to the starting base directory.
eg. Assuming Book.xml includes Introduction/Part.xml and Part.xml includes
GettingStarted/Chapter.xml the intermediate output after XInclude processing
will look something like this:
<book xml:base="Book.xml">
<part xml:base="Introduction/Part.xml">
<chapter xml:base="Introduction/GettingStarted/Chapter.xml">
<!-- stuff here, maybe a mediaobject or something -->
</chapter>
</part>
</book>
However, the DocBook v5 "relative-uri" template in "common/common.xsl" will
recursively join the directory components of the path to compute relative
paths for mediaobjects with filerefs. Suppose a fileref of
"images/SomeImage.png" is used inside Chapter.xml, the resulting path
DocBook computes will look like
"Introduction/Introduction/GettingStarted/images/SomeImage.png".
Note the redundant "Introduction" path segment here.
I don't know enough about XInclude to tell whether the behavior of the
Xerces processor is correct or whether DocBook is making an invalid
assumption.
In any case, I disabled the recursive traveral with the following temporary
patch, and now everything is fine:
Index: docbook-xsl-ns/common/common.xsl
===================================================================
--- docbook-xsl-ns/common/common.xsl (revision 1055)
+++ docbook-xsl-ns/common/common.xsl (working copy)
@@ -1722,6 +1722,11 @@
<!-- Recursively resolve xml:base attributes, up to a
full path with : in uri -->
+ <!-- Disabled by Jeff Brown on 7/18/2008 because it turns out that
+ the Xerces XInclude processor encodes xml:base using the full
+ path relative to the starting directory rather than using
+ partial paths relative to the containing element's xml:base
+ as is assumed by this code.
<xsl:if test="$base.elem/ancestor::[EMAIL PROTECTED]:base != ''] and
not(contains($base.elem/@xml:base, ':'))">
<xsl:call-template name="xml.base.dirs">
@@ -1729,6 +1734,7 @@
select="$base.elem/ancestor::[EMAIL PROTECTED]:base !=
''][1]"/>
</xsl:call-template>
</xsl:if>
+ -->
<xsl:call-template name="getdir">
<xsl:with-param name="filename" select="$base.elem/@xml:base"/>
</xsl:call-template>
Cheers,
Jeff.
P.S. Nice work on DocBook!
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
