Re: [xsl] Better include them in the XSLT 2.0 spec (Was: Re: [xsl] Time for an exslt for 2.0?)

[Colin Paul Adams <colin@xxxxxxxxxxxxxxxxxx>]

>>>>> "Dimitre" == Dimitre Novatchev <dnovatchev@xxxxxxxxx> writes:

    Dimitre> This is the problem it shouldn't, according to the XSLT
    Dimitre> 2.0 spec: there are cases when my:f($x) is my:f($x)

    Dimitre> evaluates to false().

    Dimitre> This happens when my:f() is defined in such a way that it
    Dimitre> creates new node(s) on every evaluation.

Oh dear. Yes, of course.
Which is why you would like a memoization attribute within the spec?
To clearly mark those functions which are not pure? Or to change the
semantics of functions that call xsl:element, for instance? Surely you
can't expect the WG to agree to such a substantial change at this late stage?

Or are you requesting banning non-pure functions altogether?
This would mean disallowing calls to extension functions, or else
insisting extension functions must not have side-effects. The latter
condition would be incompatible with XSLT 1.0, I think

-- 
Colin Adams
Preston Lancashire