John Burski <john(dot)burski(at)911ep(dot)com> writes:
> Could someone clarify the concept that underlies the opaque data type?
It *is* confusing, because OPAQUE is actually used for several distinct
purposes. If you look in the pg-hackers mail list archives, you'll find
discussions (as yet unimplemented) to replace OPAQUE by several
pseudo-types that have crisper meanings.
Offhand I can recall these uses:
* Sometimes OPAQUE means "this function doesn't return anything at all",
similarly to "returns void" in C and some other languages.
* Sometimes OPAQUE means "this value can be any datatype at all" (eg,
input of the count() aggregate).
* Sometimes OPAQUE means "I know exactly what this should be, but it's
an internal datatype with no SQL-level equivalent". This is usually
what's at stake with a trigger function.
* Sometimes OPAQUE means "I know exactly what this should be, but I
haven't declared it yet" (ie, a forward reference). This is needed
to declare I/O functions of user-defined types, since the system
requires the functions to already exist when the type is declared.
Confused yet? Basically, OPAQUE has been used as an easy out in rather
more places than it should have been :-(. Someday we'll clean this up.
regards, tom lane
In response to
pgsql-novice by date
|Next:||From: John Burski||Date: 2001-08-16 15:03:39|
|Subject: Re: Opaque data type|
|Previous:||From: John Burski||Date: 2001-08-16 13:59:48|
|Subject: Opaque data type|