- From: Alex Milowski <alex@milowski.org>
- Date: Sun, 23 Jul 2006 14:20:45 -0700
- To: public-xml-processing-model-wg@w3.org
Norman Walsh wrote: > / Alex Milowski <alex@milowski.org> was heard to say: > | I'd call this an error. The for-each is like a mini-pipeline and > | needs to have that extra input declared: > > Well, Jeni argued against that requirement. Hmm... I guess I need to be refreshed of the reasons why we wouldn't do this. > > | <p:for-each select="//chapter" ref="#valid/result" name="loop"> > | <p:declare-input port="#build-stylesheet/result" > | name="the-stylesheet"/> > | <p:declare-output port="result" name="chapter-docs"/> > | > | <p:step kind="transform" name="makehtml"/> > | <p:input port="document" ref="#loop/#matched"/> > | <!-- somehow we have to expose the matched regions as documents; > | here I'm imagining that for-each always declares a magically > | named input port which will be used for that purpose --> > | <p:input port="stylesheet" ref="the-stylesheet"/> > | <p:output port="result" ref="#loop/chapter-docs"/> > | </p:step> > | </p:for-each> > > But even in this case, does each iteration get a copy of the whole > #build-stylesheet/result ne the-stylesheet document? I would say it needs to have the equivalent of the same infoset each time. If an implementation needs to make a copy to guarantee that is true, then it should. That is, smart optimizations should be allowed. --Alex Milowski
Received on Sunday, 23 July 2006 21:20:59 UTC