Re: Grouping in the FAQ (was Re: Conflicting Node Values)

[David Carlisle <davidc@xxxxxxxxx>]

   > Steve Muench writes:
   >  > At least for the Oracle XSLT engine the:
   >  > 
   >  >     generate-id(.) = generate-id(key('tid',.)[1])
   >  > 
   >  > is slightly faster because it doesn't have to produce
   >  > two nodesets to merge and count as does the:
   >  > 
   >  >     count(.|key('tid',.))=1
   >  > 
   >  > approach. The latter is definitely shorter and the
   > 
   > Thats very interesting. Maybe I misinterpret Mike as saying it was
   > faster. 

   I did say the "union" approach might be marginally faster on Saxon, on
   thinking about it again I'm not so sure. I didn't spot that the
   generate-id() approach would only look at the first node in the result of
   key() whereas the union approach looks at all of them. Would need to measure
   to be sure.

   If people are going to start using generate-id() like this, though, I ought
   to think about speeding it up (if possible, without bloating the tree!)



I'm a bit confused as
generate-id(.) = generate-id(key('tid',.)[1])
tests if the current node is the first node returned by the yey
and
generate-id(.) = generate-id(key('tid',.))
is the same due to `take first node in node set'
semantics

but

count(.|key('tid',.))=1

tests that the key only returns one node, and that that is the current
one.

Don't you want

count(.|key('tid',.)[1])=1

which is equivalent to the generate-id tests.


David






 XSL-List info and archive:  http://www.mulberrytech.com/xsl/xsl-list