I'm trying to figure out how to use XSLT Streaming (to reduce memory usage) in a scenario that requires grouping (with an arbitrary number of groups) and summing the group. So far I haven't been able to find any examples. Here's an example XML
<?xml version='1.0' encoding='UTF-8'?>
<Data>
<Entry>
<Genre>Fantasy</Genre>
<Condition>New</Condition>
<Format>Hardback</Format>
<Title>Birds</Title>
<Count>3</Count>
</Entry>
<Entry>
<Genre>Fantasy</Genre>
<Condition>New</Condition>
<Format>Hardback</Format>
<Title>Cats</Title>
<Count>2</Count>
</Entry>
<Entry>
<Genre>Non-Fiction</Genre>
<Condition>New</Condition>
<Format>Paperback</Format>
<Title>Dogs</Title>
<Count>4</Count>
</Entry>
</Data>
In XSLT 2.0 I would use this to group by Genre, Condition and Format and Sum the counts.
<xsl:stylesheet version="2.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
<xsl:output method="text" indent="yes" />
<xsl:template match="/">
<xsl:call-template name="body"/>
</xsl:template>
<xsl:template name="body">
<xsl:for-each-group select="Data/Entry" group-by="concat(Genre,Condition,Format)">
<xsl:value-of select="Genre"/>
<xsl:value-of select="Condition"/>
<xsl:value-of select="Format"/>
<xsl:value-of select="sum(current-group()/Count)"/>
</xsl:for-each-group>
</xsl:template>
</xsl:stylesheet>
For output I would get two lines, a sum of 5 for Fantasy, New, Hardback and a sum of 4 for Non-Fiction, New, Paperback.
Obviously this won't work with Streaming because the sum accesses the whole group. I think I need to iterate through the document twice. The first time I could build a map of the groups (creating a new group if one doesn't exist yet). The second time The problem is I also need an accumulator for each group with a rule that matches the group, and it doesn't seem you can create dynamic accumulators.
Is there a way to create accumulators on the fly? Is there another/easier way to do this with Streaming?
To be able to use streamed grouping with XSLT 3.0 one option that I see is to first transform the element based data you have into attribute based data using a stylesheet like
<xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
xmlns:xs="http://www.w3.org/2001/XMLSchema"
xmlns:math="http://www.w3.org/2005/xpath-functions/math"
exclude-result-prefixes="xs math"
version="3.0">
<xsl:mode streamable="yes" on-no-match="shallow-copy"/>
<xsl:output indent="yes"/>
<xsl:strip-space elements="*"/>
<xsl:template match="Entry/*">
<xsl:attribute name="{name()}" namespace="{namespace-uri()}" select="."/>
</xsl:template>
</xsl:stylesheet>
then you can perfectly used streamed grouping (as far as a streamed group-by is possible at all, as far as I understand there will be some buffering necessary) as follows:
<xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
xmlns:xs="http://www.w3.org/2001/XMLSchema"
xmlns:math="http://www.w3.org/2005/xpath-functions/math"
exclude-result-prefixes="xs math"
version="3.0">
<xsl:mode streamable="yes"/>
<xsl:output method="text"/>
<xsl:template match="/">
<xsl:fork>
<xsl:for-each-group select="Data/Entry" composite="yes" group-by="@Genre, @Condition, @Format">
<xsl:value-of select="current-grouping-key(), sum(current-group()/@Count)"/>
<xsl:text> </xsl:text>
</xsl:for-each-group>
</xsl:fork>
</xsl:template>
</xsl:stylesheet>
I don't know whether first creating an attribute centric document is an option but I think it is better to share suggestions with code in an answer instead of trying to put them into a comment. And the answer in XSLT Streaming Chained Transform shows how to use Saxon 9 with Java or Scala to chain two streaming transformations without the need to write a temporary output file for the first transformation step.
As for doing it with copy-of on the original input format, Saxon 9.7 EE assesses the following as streamable and executes it with the right result:
<xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
xmlns:xs="http://www.w3.org/2001/XMLSchema"
xmlns:math="http://www.w3.org/2005/xpath-functions/math" exclude-result-prefixes="xs math"
version="3.0">
<xsl:mode streamable="yes"/>
<xsl:output method="text"/>
<xsl:template match="/">
<xsl:for-each-group select="copy-of(Data/Entry)" composite="yes"
group-by="Genre, Condition, Format">
<xsl:value-of select="current-grouping-key(), sum(current-group()/Count)"/>
<xsl:text> </xsl:text>
</xsl:for-each-group>
</xsl:template>
</xsl:stylesheet>
I am not sure it consumes less memory however than normal, tree based grouping. Perhaps you can measure with your real input data.
As a third alternative, to use a map as you seemed to want to do, here is an xsl:iterate example that iterates through the Entry elements, collecting the accumulated Count value in a map:
<xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
xmlns:xs="http://www.w3.org/2001/XMLSchema"
xmlns:math="http://www.w3.org/2005/xpath-functions/math"
xmlns:map="http://www.w3.org/2005/xpath-functions/map" exclude-result-prefixes="xs math map"
version="3.0">
<xsl:mode streamable="yes"/>
<xsl:output method="text"/>
<xsl:template match="/">
<xsl:iterate select="Data/Entry">
<xsl:param name="groups" as="map(xs:string, xs:integer)" select="map{}"/>
<xsl:on-completion>
<xsl:value-of select="map:keys($groups)!(. || ' ' || $groups(.))" separator=" "/>
</xsl:on-completion>
<xsl:variable name="current-entry" select="copy-of()"/>
<xsl:variable name="key"
select="string-join($current-entry/(Genre, Condition, Format), '|')"/>
<xsl:next-iteration>
<xsl:with-param name="groups"
select="
if (map:contains($groups, $key)) then
map:put($groups, $key, map:get($groups, $key) + xs:integer($current-entry/Count))
else
map:put($groups, $key, xs:integer($current-entry/Count))"
/>
</xsl:next-iteration>
</xsl:iterate>
</xsl:template>
</xsl:stylesheet>