Re: Initial review of xslt with no limits patch

Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com> writes:
> 2010/8/6 Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>:
>> I think there are issues here that we need to take a step back and think
>> about. Right now, thanks to the lack of documentation, we can probably
>> assume there are approximately zero users of the xslt_process parameter
>> feature. Once we document it that'll no longer be true. So right now
>> would be the time to reflect on whether this is a specification we
>> actually like or believe is usable; it'll be too late to change it
>> later.

> I know about one important user from Czech Republic

Well, if there actually is anybody who's figured it out, we could easily
have a backwards-compatible mode. Provide one variadic function that
acts as follows:
even number of variadic array elements -> they're names/values
one variadic array element -> parse it the old way
otherwise -> error

I wouldn't even bother with fixing the MAXPARAMS limitation for the
"old way" code, just make it work exactly as before.

> I'll propose a new kind of functions (only position parameter's
> function). My idea is simple - for functions with this mark the mixed
> and named notation is blocked.

We don't need random new function behaviors for this. Anyway your
proposal doesn't work at all for non-constant parameter names.

regards, tom lane