Re: [xsl] Saxon & xsltproc giving different output

There are several "messy" constructs that could be cleaned up. Try 
cleaning them up and see if this makes the problem go away.
                <xsl:when test='../module_impacted[@arch!=$device_to_output]'>
                <xsl:if test='../module_impacted/text()[.=$current_module]'>
                        <xsl:text></xsl:text>
                </xsl:if>

The predicate [@arch!=$device_to_output] rings alarm bells. When there 
is no @arch attribute (as in the example input), the predicate should be 
false. But users writing this often intended to write [not(@arch = 
$device_to_output)], which is true when @arch is absent. (The 
ungrammatical comment above the code, however, suggests that the 
programmer probably intended what was written).
The use of /text() is bad practice. It gives the wrong result when the 
text includes comments or processing instructions, and it can give the 
wrong result in some processors using an unnormalized DOM as input (it's 
not conformant, but some processors allow the input to contain multiple 
adjacent text nodes). The xsl:if expression should be written 
test="../module-impacted=$current-module".
That <xsl:text></xsl:text> is pretty strange. XSLT 2.0 is clear that it 
constructs a zero-length text node, which then disappears when it is 
added as a child to the document (root) node. XSLT 1.0 is much less 
clear what happens here. It's far from clear what the author of this 
code intended. As far as I can see, the xsl:if achieves nothing (the 
outcome is the same whether the condition is true or false). So there's 
something fishy here.
There's nothing here where the result is explicitly 
implementation-defined, but the code is entering some murky waters.
Michael Kay
Saxonica



On 17/12/2011 21:36, Russell Urquhart wrote:
> Thanks for your reply Michael.
>
> On Sat, Dec 17, 2011 at 07:05:46PM +0000, Michael Kay wrote:
> > There are two ways people can help you with this. Either they can check
> > the code by visual inspection and hope to find the errors that way, or
> > they can run it and debug it. You've provided far too much code for the
> > first approach to be practical, and not enough for the second.
> I suspected that I might have done that. The entire xslt is very large and does 
> a lot, and the rest works quite well. I also wasn't trying to get someone to 
> debug the whole thing.
>
> > You're also doing your best to confuse the reader by starting with an
> > anecdote about a different problem from the one you are asking for help
> > with.
> Sorry again. Just trying to provide info, to capture the context of the problem 
> as i encountered it.
>
> > There are legitimate reasons, incidentally, why two XSLT processors can
> > produce different results on the same input. This is especially true if
> > one of them is an XSLT 2.0 processor and the other is an XSLT 1.0
> > processor,
> This is what i thought. As i am using Saxon 9, and xsltproc, as i understand 
> it, is 1.0 only?
>
>
>
> > but it can also happen with two conformant XSLT 1.0
> > processors, because some aspects of behaviour (such as collating
> > sequences) are explicitly implementation-defined.
> Thinking this was a 1.0 issue, i downloaded Saxon 6.5.5, the last 1.0 processor 
> Saxon made, and tried it, but got identical results as i was getting from Saxon 
> 9.
>
> Let me try and recast my request.
>
> As the variable is defined:
>
> <xsl:variable name="is_module_in_device">
>            <xsl:for-each 
> select="//errata_section/devices_impacted[device_name=$device_to_output]">
>         <xsl:choose>
>                 <!-- Start Patch part 1(removing modules having and arch attribute 
> containing a different than the current device) -->
>                 <xsl:when test='../module_impacted[@arch!=$device_to_output]'>
>                 <xsl:if test='../module_impacted/text()[.=$current_module]'>
>                         <xsl:text></xsl:text>
>                 </xsl:if>
>                 </xsl:when>
>                 <!-- End Patch part 1 -->
>                 <xsl:otherwise>
>                 <xsl:if 
> test='../module_impacted/text()[.=$current_module]'><xsl:text>o</xsl:text></xsl:if>
>                 </xsl:otherwise>
>         </xsl:choose>
>            </xsl:for-each>
>          </xsl:variable>
>
> And the table is generated because:
>
> <xsl:if test='contains($is_module_in_device, "o")'>
>            <xsl:variable name="num_of_sections">
>         <!-- Start Patch part 3 -->
>              <xsl:value-of select="string-length($is_module_in_device) + 
> string-length($removed_modules)"/>
>         <!-- End Patch part 3 -->
>            </xsl:variable>
>
>            <table>
>
>
> How could is_module_in_device have different values for xslt processors?
>
> Thanks,
>
> Russ
>
> > On 17/12/2011 17:05, Russell Urquhart wrote:
> > > Hi,
> > >
> > > I have to support a legacy xslt that transforms an xml flle, organizes it, and 
> > > then outputs a docbook style xml file.
> > >
> > > When i took over this, i was running Saxon on the xslt, without incident. 
> > > Recently i noticted that some elements were not being caught and output. I 
> > > determined that the previous users had been using xsltproc, so i used that and 
> > > the output i was getting was in sync with what they had had.
> > >
> > > I know that Saxon is the standard, but i didn't have the time or knowledge to 
> > > discern the differecnes in the xslt.
> > >
> > > I am now experiencing a related problem, where stange values are being 
> > > generated by the xslt, while Saxon does not output any of these specific 
> > > elements.
> > >
> > > The input data, is of the following:
> > >
> > > errata_section id="i707" errata_type="caution">
> > >         <title>Adaptive Body Bias Strategy</title>
> > >      <devices_impacted>
> > >           <device_name>xxxx2345</device_name>
> > >           <device_es>1.0</device_es>
> > >           <device_es>1.1</device_es>
> > >         </devices_impacted>
> > > <devices_impacted>
> > >           <device_name>xxxx2346</device_name>
> > >           <device_es>1.0</device_es>
> > >         </devices_impacted>
> > >
> > >         <module_impacted>PRCM</module_impacted>
> > >       </errata_section>
> > >
> > > The code is called to generate a table of modules impacted, for a given device 
> > > impacted.
> > >
> > > It's called like this:
> > >
> > >         <xsl:call-template name="authoring">
> > >             <xsl:with-param name="device_to_output">xxxx2345</xsl:with-param>
> > >         </xsl:call-template>
> > >
> > > A key is created:
> > >
> > > xsl:key name="module-index" match="errata_section" use="module_impacted"/>
> > >
> > >
> > > The code in question i think is the following, with original comments in tact 
> > > (Sorry if there is unnecessary code here):
> > >
> > > xsl:template name="table_of_section_per_module">
> > >        <xsl:param name="device_to_output"/>
> > >
> > >        <xsl:element name="article">
> > >        <xsl:attribute name="id">errata_module_impacted</xsl:attribute>
> > >        <xsl:attribute name="arch"><xsl:value-of 
> > > select='$device_to_output'/></xsl:attribute>
> > >        <title>Modules Impacted</title>
> > > <!--     This para element is being deleted here because it causes an error. Para 
> > > elements cannot contain stuff.-->
> > > <!--<para>-->
> > >        <!-- A little complicated but seems to work fine ! -->
> > >        <xsl:for-each 
> > > select="//errata_section[generate-id(.)=generate-id(key('module-index', 
> > > module_impacted)[1])]">
> > >           <xsl:sort select="module_impacted"/>
> > >           <xsl:variable name="current_module"><xsl:value-of 
> > > select="module_impacted"/></xsl:variable>
> > >           <!-- The is an ugly kludge: we check each section that applies to current chip 
> > > and update the "is_module_in_device"
> > >               variable if we find it impacts the current module. In the end we 
> > > have either an empty variable (module is not in
> > >               current device) or a variable containing lots of "o" concatenated 
> > > (module is in current device) -->
> > >
> > >           <xsl:variable name="is_module_in_device">
> > >             <xsl:for-each 
> > > select="//errata_section/devices_impacted[device_name=$device_to_output]">
> > >         <xsl:choose>
> > >                 <!-- Start Patch part 1(removing modules having and arch attribute 
> > > containing a different than the current device) -->
> > >                 <xsl:when test='../module_impacted[@arch!=$device_to_output]'>
> > >                 <xsl:if test='../module_impacted/text()[.=$current_module]'>
> > >                         <xsl:text></xsl:text>
> > >                 </xsl:if>
> > >                 </xsl:when>
> > >                 <!-- End Patch part 1 -->
> > >                 <xsl:otherwise>
> > >                 <xsl:if 
> > > test='../module_impacted/text()[.=$current_module]'><xsl:text>o</xsl:text></xsl:if>
> > >                 </xsl:otherwise>
> > >         </xsl:choose>
> > >             </xsl:for-each>
> > >           </xsl:variable>
> > >         
> > >         <!-- Start Patch part 2 -->
> > >         <xsl:variable name="removed_modules">
> > >             <xsl:for-each 
> > > select="//errata_section/devices_impacted[device_name=$device_to_output]">
> > >         <xsl:choose>
> > >                 <xsl:when test='../module_impacted[@arch!=$device_to_output]'>
> > >                 <xsl:if test='../module_impacted/text()[.=$current_module]'>
> > >                         <xsl:text>r</xsl:text>
> > >                 </xsl:if>
> > >                 </xsl:when>
> > >         </xsl:choose>
> > >             </xsl:for-each>
> > >           </xsl:variable>
> > >         <!-- End Patch part 2 -->
> > >         
> > >           <xsl:if test='contains($is_module_in_device, "o")'>
> > >             <xsl:variable name="num_of_sections">
> > >         <!-- Start Patch part 3 -->
> > >               <xsl:value-of select="string-length($is_module_in_device) + 
> > > string-length($removed_modules)"/>
> > >         <!-- End Patch part 3 -->
> > >             </xsl:variable>
> > >
> > >             <table>
> > >              <xsl:attribute name="id">
> > >              <xsl:value-of select="generate-id(module_impacted)"/>
> > >              </xsl:attribute>
> > >               <title>
> > >               <xsl:text>Module</xsl:text>
> > >               <xsl:value-of select="module_impacted"/>
> > >               <xsl:text>   (</xsl:text><xsl:value-of 
> > > select="$num_of_sections"/><xsl:text>   section</xsl:text>
> > >               <!-- a little grammar ! -->
> > >               <xsl:if 
> > > test="$num_of_sections&gt;1"><xsl:text>s</xsl:text></xsl:if><xsl:text>)</xsl:text>
> > >               </title>
> > >               <xsl:element name="tgroup">
> > >               <xsl:attribute name="cols">
> > >               <xsl:choose>
> > >                 <xsl:when test='$show_review="yes"'>3</xsl:when>
> > >                 <xsl:otherwise>2</xsl:otherwise>
> > >               </xsl:choose>
> > >               </xsl:attribute>
> > >               <xsl:element name="colspec">
> > >                 <xsl:attribute name="colwidth">1*</xsl:attribute>
> > >                 <xsl:attribute name="colname">_1</xsl:attribute>
> > >               </xsl:element>
> > >               <xsl:choose>
> > >                   <xsl:when test='$show_review="yes"'>
> > >                      <xsl:element name="colspec">
> > >                         <xsl:attribute name="colwidth">6*</xsl:attribute>
> > >                         <xsl:attribute name="colname">_2</xsl:attribute>
> > >                      </xsl:element>
> > >                      <xsl:element name="colspec">
> > >                         <xsl:attribute name="colwidth">1*</xsl:attribute>
> > >                         <xsl:attribute name="colname">_3</xsl:attribute>
> > >                      </xsl:element>
> > >                   </xsl:when>
> > >                   <xsl:otherwise>
> > >                      <xsl:element name="colspec">
> > >                         <xsl:attribute name="colwidth">8*</xsl:attribute>
> > >                         <xsl:attribute name="colname">_2</xsl:attribute>
> > >                      </xsl:element>
> > >                   </xsl:otherwise>
> > >               </xsl:choose>
> > >               <tbody>
> > >               <row>
> > >               <xsl:element name="entry">
> > >               <xsl:attribute name="morerows"><xsl:value-of 
> > > select="$num_of_sections"/></xsl:attribute>
> > >               <xsl:value-of select="module_impacted"/>
> > >               </xsl:element>
> > >               <entry>Section</entry>
> > >               <xsl:if test='$show_review="yes"'>
> > >               <entry>Review status</entry>
> > >               </xsl:if>
> > >               </row>
> > >                 <xsl:for-each select="key('module-index', module_impacted)">
> > >                   <xsl:if 
> > > test="devices_impacted/device_name/text()[.=$device_to_output]">
> > >                     <row>
> > >                     <entry>
> > >                 <xsl:value-of select="@id" /><xsl:text>:</xsl:text>
> > >                     <xsl:element name="xref">
> > >                     <xsl:attribute name="linkend">sect_<xsl:value-of select="@id" 
> > > />
> > >                     </xsl:attribute></xsl:element>
> > >                     </entry>
> > >                     <xsl:if test='$show_review="yes"'>
> > >                     <entry>
> > >                     <xsl:value-of select="review_status" />
> > >                     </entry>
> > >                     </xsl:if>
> > >                      </row>
> > >
> > >
> > > My issue is, Saxon doesn't find ALL the modules impacted for a given device. 
> > > Xsltproc, in the past has found all moduls impacted for a given device, but, 
> > > recently, has incorrectly found unrelated modules impacted, for a given device.
> > >
> > > I suspect, by their own admission, that the code at this point is kludgey, and 
> > > they are taking advantage of some aspect of how xsltproc handles xslt 1.0 style 
> > > code?
> > >
> > > Please any help on this would be great. I want to only have to use Saxon to do 
> > > this transform!
> > >
> > > Thanks again,
> > >
> > > Russ
> > >
> > > --~------------------------------------------------------------------
> > > XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list
> > > To unsubscribe, go to: http://lists.mulberrytech.com/xsl-list/
> > > or e-mail:<mailto:xsl-list-unsubscribe(_at_)lists(_dot_)mulberrytech(_dot_)com>
> > > --~--
> > --~------------------------------------------------------------------
> > XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list
> > To unsubscribe, go to: http://lists.mulberrytech.com/xsl-list/
> > or e-mail:<mailto:xsl-list-unsubscribe(_at_)lists(_dot_)mulberrytech(_dot_)com>
> > --~--
> --~------------------------------------------------------------------
> XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list
> To unsubscribe, go to: http://lists.mulberrytech.com/xsl-list/
> or e-mail:<mailto:xsl-list-unsubscribe(_at_)lists(_dot_)mulberrytech(_dot_)com>
> --~--

--~------------------------------------------------------------------
XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list
To unsubscribe, go to: http://lists.mulberrytech.com/xsl-list/
or e-mail: <mailto:xsl-list-unsubscribe(_at_)lists(_dot_)mulberrytech(_dot_)com>
--~--