[XSL-LIST Mailing List Archive Home] [By Thread] [By Date]

Re: [xsl] Re: The Solution -- Re: how to rearrange nodes based on a dependency graph?


Subject: Re: [xsl] Re: The Solution -- Re: how to rearrange nodes based on a dependency graph?
From: Gunther Schadow <gunther@xxxxxxxxxxxxxxxxxxxxxx>
Date: Fri, 21 Dec 2001 13:04:14 -0500

As beautifully as this is designed, the only sad thing is that
it can't work without using some Microsoft (or other) extension.
Apparently the node-set function is available in many xslt
processors, but then why is it not part of the specification?

It is my reading, correctme if I'm wrong, that the only reason
why this node-set function needs to exist in the first place is
because the result tree fragment has some ugly restriction. The
XSLT spec mumbles about it being only accessible through string
functions etc. The node-set apparently turns that result tree
fragment into real XPath-inspectable trees.

Furthermore, the only place where this is really needed is for
the concatenation of the node sets from pSorted and vNextLevel.
That should have been easily possible in basic XPath. I tried
rewriting:

      <xsl:variable name="vrtfNewSorted">
        <xsl:copy-of select="$pSorted"/>
        <xsl:copy-of select="$vNextLevel"/>
      </xsl:variable>

to this:

      <xsl:variable name="XvrtfNewSorted"
                    select="$pSorted|$vNextLevel"/>

but apparently this orders things in the document order, not
in the order specified. It's weird that the XPath spec talks
about node sets when in reality they are ordered. And all of
this only to prevent the user from rearranging the ordering.

regards
-Gunther


Gunther Schadow wrote:


Wow, thanks Dimitre, this looks exactly right! And so much more
elegant than what I have been fiddling with! I had been close
to banging my "nodes-seen" list into XSL but your solution is
so much better!

Thanks,
-Gunther

Dimitre Novatchev wrote:

Hi Gunther,

If I'm not wrong, this problem is a case of the "topological sort" problem.

The solution bellow arranges the DAG into levels of hierarchy -- the nodes that do
not depend on other nodes are at the top, the nodes that depend ***only*** on the
nodes in the sorted hierarchy (up to this moment) form the next level of the
hierarchy.


Suppose that we have the following source xml document:

<document xml:space="preserve">
    <frag id='1'>
        <requires>2</requires>
        <requires>3</requires>
    </frag>

    <frag id='2'>
        <requires>4</requires>
        <requires>5</requires>
    </frag>

    <frag id='3'>
        <requires>5</requires>
    </frag>

<frag id='4' />

    <frag id='5' />
</document>

The hierarchical representation is:

      1
     / \
    2   3
   / \ /
  4   5


And here's the stylesheet:


<xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
xmlns:msxsl="urn:schemas-microsoft-com:xslt">
<xsl:output indent="yes" omit-xml-declaration="yes"/>
<xsl:template match="/">
<xsl:call-template name="topSort">
<xsl:with-param name="pSorted" select="/*/frag[not(requires)]"/>
<xsl:with-param name="pUnsorted" select="/*/frag[requires]"/>
</xsl:call-template>
</xsl:template>
<xsl:template name="topSort">
<xsl:param name="pSorted" select="/.."/>
<xsl:param name="pUnsorted" select="/.."/>


<xsl:variable name="vNextLevel"
select="$pUnsorted[not(requires[not(. = $pSorted/@id)])]"/>
<xsl:choose>
<xsl:when test="not($vNextLevel)">
<xsl:copy-of select="$pSorted"/>
</xsl:when>
<xsl:otherwise>
<xsl:variable name="vrtfNewSorted">
<xsl:copy-of select="$pSorted"/>
<xsl:copy-of select="$vNextLevel"/>
</xsl:variable>


<xsl:variable name="vCntNextLevel" select="count($vNextLevel)"/>
<xsl:variable name="vNewUnsorted"
select="$pUnsorted[not(@id = $vNextLevel/@id)] "/>
<xsl:call-template name="topSort">
<xsl:with-param name="pSorted" select="msxsl:node-set($vrtfNewSorted)/*"/>
<xsl:with-param name="pUnsorted" select="$vNewUnsorted"/>
</xsl:call-template>


      </xsl:otherwise>
    </xsl:choose>
  </xsl:template>
</xsl:stylesheet>

When applied on the above xml source document, the result of the transformation is:

<frag id="4" />
<frag id="5" />
<frag id="2">
        <requires>4</requires>
        <requires>5</requires>
</frag>
<frag id="3">
        <requires>5</requires>
</frag>
<frag id="1">
        <requires>2</requires>
        <requires>3</requires>
</frag>


The tricky part is to express in a single XPath expression the nodes that will form
the next level:


<xsl:variable name="vNextLevel"
              select="$pUnsorted[not(requires[not(. = $pSorted/@id)])]"/>

This is: all elements from $pUnsorted, which do not have any "requires" child that
is not equal to the "id" of one of the elements in the sorted hierarchy $pSorted.


Hope that this really helped.

Cheers,
Dimitre Novatchev.



Gunther Schadow <gunther at aurora dot regenstrief dot org> wrote:
--------------------------------------------------------------------------------



Hi XSL listers,


I'm hung up on a puzzling problem and need your help. I am
relatively new to XSL but thanks to the pretty readable
specification and some example I was so far able to find
my way. Except when I wanted to be really clever, like here.

The purpose of what I'm doing is kind of what you might
know as "literal programming." It's writing a document
that explains the detail of some formal language utterance,
such as a program, or an XML schema or an IDL specification
or whatever have you. The document is to be written for
readability and may well have a different flow than what's
required for the formal language (program or xsd or whatever.)
For example, in describing a Pascal program, I might want to
discuss an outline of the main program first before I go
into the detail of the procedures. Yet in the program text,
the procedures need to come before the main program. The
point of course is that the text document should be the main
focus of development and maintenance, and the program text
should be generated from there. That's where XSLT comes
in and falls short? (you tell me if it does of if I fall
short :-)

Here is an example. This XML instance discusses a mocked up
up XML schema for defining some data.


<document>


Blah bla

<frag id='1' requires='2'>
    BEGIN
      DoSomethingUseful();
    END
</frag>

Blah blah

<frag id='2'>
   PROCEDURE DoSomethingUseful
   BEGIN
     ...
   END
</frag>

</document>




The <frag> tag has an 'id' (ID) attribute with matching 'requires' (IDREFS) attribute.

The XSL templates will go through the document and discard
everything, except when encountering a fragment. It will
then will hunt down the requires idrefs to find other
fragments that should come first. Of course, if a fragment has
already been emitted, it shouldn't appear again. And that't
really my big problem.

Here is my XSL so far:

<xsl:key name='fragkey' match='frag' use='@id' />

<xsl:template match='frag'>
   <xsl:for-each select='@requires'>
     <xsl:apply-templates select="key('fragkey',.)"/>
   </xsl:for-each>

   <xsl:for-each select='*'>
     <xsl:call-template name='copy-stuff'/>
   </xsl:for-each>
</xsl:template>

<xsl:template name='copy-stuff'>
   <xsl:copy>
     <xsl:for-each select='@*|*'>
       <xsl:call-template name='copy-stuff'/>
     </xsl:for-each>
   </xsl:copy>
</xsl:template>



This works nicely, except for the fact that fragment number 2
is emitted twice, first as the dependency of fragment 1 and
then because it appears as the next fragment in the document.
(Similarly this runs amok if there's a circular dependency.)

I tried to use a variable as a check-off list of fragments
already emitted, but that's not possible because XSL "variables"
are actually constants. When I try to write such a function
in LISP without using global variables and side effects, it
get's pretty difficult and the only way to survive here is
because I can examine return values. In XSL, I cannot examine
the output tree, or can I?

Any suggestions are appreciated. Of course I can dumb down
this system and instead of requirement links I could use
some kind of sequence number or forward linking and sort by
that etc. But I feel that the dependency graph is the most
appropriate way to represent this and if XSL is a complete
functional language it should have some way to deal with that.

regards,
-Gunther




__________________________________________________ Do You Yahoo!? Check out Yahoo! Shopping and Yahoo! Auctions for all of your unique holiday gifts! Buy at http://shopping.yahoo.com or bid at http://auctions.yahoo.com





--
Gunther Schadow, M.D., Ph.D.                    gschadow@xxxxxxxxxxxxxxx
Medical Information Scientist      Regenstrief Institute for Health Care
Adjunct Assistant Professor        Indiana University School of Medicine
tel:1(317)630-7960                         http://aurora.regenstrief.org



XSL-List info and archive: http://www.mulberrytech.com/xsl/xsl-list



Current Thread