Re: Making vanilla implementation of position() DTRT

/ ht@inf.ed.ac.uk (Henry S. Thompson) was heard to say:
[...Henry's cogent arguments elided...]

Looking at the example in Henry's mail, it suddenly became clear to me
why we can't use position() inside a for-each or viewport as we've
been considering. The observation is so simple, I can't believe no one
else made it before, so let me know if I'm once again overlooking the
obvious.

Consider his example:

 <p:for-each name="myloop">
   <p:group>
     <p:option name="index" select="position()"/>
      . . .

What is the context for that p:option? Depending on what we say about
the default context for p:option, it's equivalent to one of the
following two rewrites:

   <p:option name="index" select="position()">
     <p:empty/>
   </p:option>

or

   <p:option name="index" select="position()">
     <p:pipe step="myloop" port="current"/>
   </p:option>

In the former case, the expression is a dynamic error. In the latter
case, position() = 1 because current doesn't produce a sequence.

I don't think there's any existing XProc construct which will evaluate
the XPath expression that contains position() in the right context.

So I think we're stuck with p:index('stepname') for counting
iterations in p:for-each and p:viewport.

Henry's analysis does make it clear, I think, that position() is the
right answer for atomic steps that accept an XPath expression.

                                        Be seeing you,
                                          norm

-- 
Norman Walsh <ndw@nwalsh.com> | Everything should be made as simple as
http://nwalsh.com/            | possible, but no simpler.