XSLT: Merge a set of tree hierarchies - xslt

Question

I have an XML document based what Excel produces when saving as "XML Spreadsheet 2003 (*.xml)".
The spreadsheet itself contains a header section with a hierarchy of labels:
| A B C D E F G H I
-+-----------------------------------------------------
1| a1 a2
2| a11 a12 a13 a21 a22
3| a111 a112 a121 a122 a131 a132 a221 a222
This hierarchy is present on all sheets in the workbook, and looks more or less the same everywhere.
Excel XML works exactly like ordinary HTML tables. (<row>s that contain <cell>s). I have been able to transform everything into such a tree structure:
<node title="a1" col="1">
<node title="a11" col="1">
<node title="a111" col="1"/>
<node title="a112" col="2"/>
</node>
<node title="a12" col="3">
<node title="a121" col="3" />
<node title="a122" col="4" />
</node>
<!-- and so on -->
</node>
But here is the complication:
there is more than one worksheet, so there is a tree for each of them
the hierarchy may be slightly different on each sheet, the trees will not be equal (for example, sheet 2 may have "a113", while the others don't)
tree depth is not explicitly limited
the labels however are meant to be the same across all sheets, which means they can be used for grouping
I'd like to merge these separate trees into one that looks like this:
<node title="a1">
<col on="sheet1">1</col>
<col on="sheet2">1</col>
<node title="a11">
<col on="sheet1">1</col>
<col on="sheet2">1</col>
<node title="a111">
<col on="sheet1">1</col>
<col on="sheet2">1</col>
</node>
<node title="a112">
<col on="sheet1">2</col>
<col on="sheet2">2</col>
</node>
<node title="a113"><!-- different here -->
<col on="sheet2">3</col>
</node>
</node>
<node title="a12">
<col on="sheet1">3</col>
<col on="sheet2">4</col>
<node title="a121">
<col on="sheet1">3</col>
<col on="sheet2">4</col>
</node>
<node title="a122">
<col on="sheet1">4</col>
<col on="sheet2">5</col>
</node>
</node>
<!-- and so on -->
</node>
Ideally I'd like to be able to do the merge before I even build the three structure from the Excel XML (if you get me started on this, it'd be great). But since I have no idea how I would do this, a merge after the trees have been built (i.e.: the situation described above) will be fine.
Thanks for your time. :)

Answer 1

Here is one possible solution in XSLT 1.0:
<xsl:stylesheet version="1.0"
xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
<xsl:output omit-xml-declaration="yes" indent="yes"/>
<xsl:template match="/*">
<t>
<xsl:apply-templates
select="node[#title='a1'][1]">
<xsl:with-param name="pOther"
select="node[#title='a1'][2]"/>
</xsl:apply-templates>
</t>
</xsl:template>
<xsl:template match="node">
<xsl:param name="pOther"/>
<node title="{#title}">
<col on="sheet1">
<xsl:value-of select="#col"/>
</col>
<xsl:choose>
<xsl:when test="not($pOther)">
<xsl:apply-templates mode="copy">
<xsl:with-param name="pSheet" select="'sheet1'"/>
</xsl:apply-templates>
</xsl:when>
<xsl:otherwise>
<col on="sheet2">
<xsl:value-of select="$pOther/#col"/>
</col>
<xsl:for-each select=
"node[#title = $pOther/node/#title]">
<xsl:apply-templates select=".">
<xsl:with-param name="pOther" select=
"$pOther/node[#title = current()/#title]"/>
</xsl:apply-templates>
</xsl:for-each>
<xsl:apply-templates mode="copy" select=
"node[not(#title = $pOther/node/#title)]">
<xsl:with-param name="pSheet" select="'sheet1'"/>
</xsl:apply-templates>
<xsl:apply-templates mode="copy" select=
"$pOther/node[not(#title = current()/node/#title)]">
<xsl:with-param name="pSheet" select="'sheet2'"/>
</xsl:apply-templates>
</xsl:otherwise>
</xsl:choose>
</node>
</xsl:template>
<xsl:template match="node" mode="copy">
<xsl:param name="pSheet"/>
<node title="{#title}">
<col on="{$pSheet}">
<xsl:value-of select="#col"/>
</col>
<xsl:apply-templates select="node" mode="copy">
<xsl:with-param name="pSheet" select="$pSheet"/>
</xsl:apply-templates>
</node>
</xsl:template>
</xsl:stylesheet>
When the above transformation is applied on this XML document (the concatenation of the two XML documents under a common top node -- left as an exercise for the reader :) ):
<t>
<node title="a1" col="1">
<node title="a11" col="1">
<node title="a111" col="1"/>
<node title="a112" col="2"/>
</node>
<node title="a12" col="3">
<node title="a121" col="3" />
<node title="a122" col="4" />
</node>
<!-- and so on -->
</node>
<node title="a1" col="1">
<node title="a11" col="1">
<node title="a111" col="1"/>
<node title="a112" col="2"/>
<node title="a113" col="3"/>
</node>
<node title="a12" col="4">
<node title="a121" col="4" />
<node title="a122" col="5" />
</node>
<!-- and so on -->
</node>
</t>
The wanted result is produced:
<t>
<node title="a1">
<col on="sheet1">1</col>
<col on="sheet2">1</col>
<node title="a11">
<col on="sheet1">1</col>
<col on="sheet2">1</col>
<node title="a111">
<col on="sheet1">1</col>
<col on="sheet2">1</col>
</node>
<node title="a112">
<col on="sheet1">2</col>
<col on="sheet2">2</col>
</node>
<node title="a113">
<col on="sheet2">3</col>
</node>
</node>
<node title="a12">
<col on="sheet1">3</col>
<col on="sheet2">4</col>
<node title="a121">
<col on="sheet1">3</col>
<col on="sheet2">4</col>
</node>
<node title="a122">
<col on="sheet1">4</col>
<col on="sheet2">5</col>
</node>
</node>
</node>
</t>
Do note the following:
We suppose that both top node elements have "a1" as the value of their title attribute. This can easily be generalized.
The template matching node has a parameter named pOther, which is the corresponding element named node from the other document. This template is applied - to only if $pOther exists.
When no corresponding element named node exists, another template, also matching node, but in mode copy is applied. This template has a parameter named pSheet, the value of which is the sheet name (string) this element belongs to.

Answer 2

How about a callable template taking the sheet number as a parameter, which examines the input and returns the correct "col" node if it appears in that sheet's XML, and nothing if it doesn't. At each node, call it once for each sheet.
To merge the trees, maybe a template that looks for all children of the current node in any sheet, and recurses on itself for each of them.
Sorry no sample code, I find writing XSLT to be pretty slow, probably because I don't do it often. So I may well have missed something crucial. But putting it all together would give something like:
get the title of "/node". With that title:
search all sheets for this title, emitting the "col" node for each
search all sheets for children of nodes with this title (discarding duplicates)
recurse on each of those titles.
Here are some snippets for removing duplicates in various ways:
http://www.dpawson.co.uk/xsl/sect2/N2696.html
Reading multiple documents is processor-dependent, but if all else fails a bit of cut-and-pastery with any old scripting language would probably do, provided that you know they'll all have the same encoding, don't use conflicting ids, and so on.