Re: pg_type.h regression?

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: "Greg Sabino Mullane" <greg(at)turnstep(dot)com>
Cc: pgsql-hackers(at)postgresql(dot)org
Subject: Re: pg_type.h regression?
Date: 2008-09-24 02:16:34
Message-ID: 16959.1222222594@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

"Greg Sabino Mullane" <greg(at)turnstep(dot)com> writes:
>> DefineType currently always sets the array's typdelim to ',', so I
>> figured that the built-in types should match that. But we could
>> easily change both of them to do the other. Thoughts?

> I'd slightly lean towards keeping it the way it has been (semicolon in both),
> since it doesn't matter to the backend, and who knows what else it may break.

Well, the problem is that we'd *not* be keeping it the same for
user-defined types. But the odds that that's really an issue are
admittedly small.

I thought of an argument for changing to array-uses-element's-typdelim,
though. If we ever go over to a scheme where arrays are one-dimensional
and you build up N-dimensional arrays using arrays of array objects
(where the outer array treats the inner as a black box), then we would
need the typdelims to match to preserve the current I/O behavior.
I'm not really enamored of making such a change myself, but I seem to
recall a couple of people advocating for it.

So, if no objections I'll revert the change to _box's typdelim and
change DefineType to copy the element typdelim to the array type.

regards, tom lane

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Andrew Dunstan 2008-09-24 02:17:46 Re: [HACKERS] 0x1A in control file on Windows
Previous Message Tom Lane 2008-09-24 01:58:36 Re: Toasted table not deleted when no out of line columns left