[xsl] Re: xslt 3.0 possible feature - some sort of collection to help when streaming

Subject: [xsl] Re: xslt 3.0 possible feature - some sort of collection to help when streaming
From: "Vladimir Nesterovsky" <vladimir@xxxxxxxxxxxxxxxxxxxx>
Date: Fri, 11 Mar 2011 09:08:28 +0200
xsl:iterate works but it's limited to a single level of iteration, as
this way one cannot easily collect output from nested transformations.

I read Andrew Welch's request as a wish to have
a way to produce multiple intermediate outputs at once.

Indeed, during transformation you can consume many (intermediate) inputs.
It would be symmetric to expect multiple (intermediate) outputs.
--
Vladimir Nesterovsky
http://www.nesterovsky-bros.com/

The approach in the spec to this kind of requirement is the xsl:iterate instruction. Instead of using xsl:message to output the data that you want to process at the end, you append it to a sequence which is passed as a parameter to each iteration of xsl:iterate, and it's then available for processing at the end using xsl:on-completion. We're adding support for maps to the language with this use case very much in mind; it makes it easier to accumulate more complex data on your single streamed pass through the input.

Michael Kay
Saxonica


On 10/03/2011 18:28, Andrew Welch wrote:
Hi,

Often we 'hack' xsl:message to output some information to post process
later, and it struck me that it would be useful to have a way of
accessing what has been sent to xsl:message during the same transform.
  This would be especially useful when streaming, as processing the
data twice for different purposes is less than ideal.

For example, you for-each over some data but want to also do something
with @names later:

<xsl:for-each select="/millions/and/millions">

[ process each item ]

   <xsl:message select="concat('Processed ', @name)"/>
</xsl:for-each>

then later in the same template, say to create an index or list of links:

<xsl:for-each select="get-messages()">
[ process the items stored earlier 'Processed name1' , 'Processed name2' etc ]


Is there anything conceptually wrong with this, other than the
non-guaranteed order of processing?  If not, then it could be a proper
instruction rather than xsl:message, eg a top-level element to set it
up:

<xsl:appender name="someAppender" option1="foo" option2="bar"/>

with:

<xsl:append name="someAppender"> text to append</xsl:append>

and then to access it:

get-appender('someAppender')

Current Thread