On Sun, Aug 9, 2009 at 2:17 PM, Tom Lane<tgl@sss.pgh.pa.us> wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
>> On Sun, Aug 9, 2009 at 12:27 PM, Tom Lane<tgl@sss.pgh.pa.us> wrote:
>>> Now that I've started to read this patch ... exactly what is the
>>> argument for allowing a "mixed" notation (some of the parameters named
>>> and some not)? ISTM that just serves to complicate both the patch
>>> and the user's-eye view, for no real benefit.
>
>> Wow, I can't imagine not supporting that. Doesn't every language that
>> supports anything like named parameters also support a mix?
>
> I don't know if that's true, [...]
I'm fairly sure it is.
> and I definitely don't know if they all
> handle corner cases identically.
I'm 100% positive that they don't.
> I think this patch is an exercise in
> guessing at what the SQL committee will eventually do, and as such, we
> should avoid like the plague making any guesses that carry significant
> risk of being semantically incompatible with what they eventually do.
> The risk/reward ratio isn't good enough.
I completely agree; I would have chosen to boot the entire patch for
exactly that reason. However, given that you don't seem to want to do
that, I was just offering my thoughts on the various possible methods
of eliminating that risk, since the one you're suggesting doesn't seem
ideal to me. I thought that perhaps providing some examples of how
other programming languages handle this problem might be helpful - in
particular, I think the Lisp model, where each parameter must be
EITHER named or positional, has a lot to recommend it.
There is little doubt here that what gets committed here will be what
you choose to commit.
...Robert
