Re: [xsl] streaming XSLT creating a header from a first record

Subject: Re: [xsl] streaming XSLT creating a header from a first record
From: "Geert Bormans geert@xxxxxxxxxxxxxxxxxxx" <xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx>
Date: Tue, 8 May 2018 15:58:34 -0000
Interesting Martin, thanks, will give that a go

Best regards,

Geert Bormans

--------------------------------------------------------------
Markup UK - a conference about XML and other mark-up languages
London, June 9b10 2018
Programme now available at [ http://markupuk.org/speakers.xhtml |
http://markupuk.org/speakers.xhtml ]
Early bird registration extended [ http://markupuk.org/registration.xhtml |
http://markupuk.org/registration.xhtml ]

----- Oorspronkelijk bericht -----
Van: "Abel Braaksma, (Exselt) abel@xxxxxxxxxx"
<xsl-list-service@xxxxxxxxxxxxxxxxxxxxxx>
Aan: "xsl-list" <xsl-list@xxxxxxxxxxxxxxxxxxxxxx>
Verzonden: Dinsdag 8 mei 2018 17:55:05
Onderwerp: Re: [xsl] streaming XSLT creating a header from a first record

On 08.05.2018 17:33, Martin Honnen martin.honnen@xxxxxx wrote:
> On 08.05.2018 17:19, Geert Bormans geert@xxxxxxxxxxxxxxxxxxx wrote:
>> All,
>>
>> I have a potentially huge tabular structure in an XML file: one
>> "table'' element, many "row" elements, each having a series of attributes
>>
>> I want to process them using a streaming XSLT, but I am hitting an issue
>>
>> I want to use information from the first row in a header construct, in
>> a way using the first row information twice
>> That is fairly easy done if the result were something like CSV,
>> because then I can deal with the header and first row in one go
>> but now I need the header in some sort of metadata section at the
>> start of the document
>> and all the rows including row one in an adjacent structure, but
>> wrapped (the results of the row processing in a container)
>>
>> I tried to create the full structure in a template for the first row
>> and process the following siblings of the first row... but that can't
>> be done "when the context posture is striding" (sic)
>>
>> Any ideas? Thanks,
>
> How about setting up an accumulator
>
> <xsl:accumulator name="first-row-values" as="map(xs:string, xs:string)"
> initial-value="map{}" streamable="yes">
>  B  <xsl:accumulator-rule match="table/row[1]"
>  B B B B  select="map:merge(@*!map { node-name() : data(.) })"/>
> </xsl:accumulator>
>
> then you can use that map in all following rows.


Hm, I forgot that you can't have positional predicates in a streamable
match pattern so it seems you first need to set up an additional
accumulator counting the rows and then to make sure the other
accumulator only stores the attribute values for the first row.
Additionally I had the wrong map type, so

	<xsl:mode streamable="yes" on-no-match="shallow-copy"
use-accumulators="#all"/>


	<xsl:accumulator name="row-count" as="xs:integer" streamable="yes"
initial-value="0">
		<xsl:accumulator-rule match="table/row" select="$value + 1"/>
	</xsl:accumulator>

	<xsl:accumulator name="first-row-values" as="map(xs:QName, xs:string)"
initial-value="map{}" streamable="yes">
		<xsl:accumulator-rule match="table/row"
			select="if (accumulator-before('row-count') eq 1)
			        then map:merge(@*!map { node-name() : string(.) })
			        else $value"/>
	</xsl:accumulator>


should work better. Only drawback is that maps don't preserve the order
of entries although as the data comes from XML attributes which aren't
ordered either that might not matter.

Current Thread