Re: need of anonymous record

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Craig Ringer <craig(at)2ndquadrant(dot)com>
Cc: Peter Krauss <ppkrauss(at)gmail(dot)com>, pgsql-hackers(at)postgresql(dot)org
Subject: Re: need of anonymous record
Date: 2014-05-04 14:30:57
Message-ID: 23777.1399213857@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Craig Ringer <craig(at)2ndquadrant(dot)com> writes:
> What you appear to want is to access arbitrary fields of a record by name.

> The reason this isn't supported directly in PL/PgSQL is, AFAIK, mainly
> an issue of data typing. Each field in a record may be of a different
> type. So the return type would change based on the field accessed at
> execution time, which isn't something PL/PgSQL or regular SQL likes very
> much, as plans are formed and cached in part based on the types involved.

Right. For better or worse, SQL is a strongly-typed language. PG's
emphasis on datatype extensibility makes this constraint even tighter,
since you can't even tell e.g. which operators or functions would be
valid to apply to a field of indeterminate type.

> We could and probably should provide an operator or function to access
> records of a field in their text form by name. Patches are welcome ;-)

As was noted upthread, if you're satisfied with pretending all fields
of a record are text, it's not clear why you wouldn't be happy using
hstore instead. Or JSON, which at least has heard of numbers as well
as text. I think the current push to improve json/jsonb is driven
largely by the desire to handle scenarios of this sort.

regards, tom lane

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Stephen Frost 2014-05-04 15:03:56 Re: pgaudit - an auditing extension for PostgreSQL
Previous Message Andrew Dunstan 2014-05-04 14:24:54 Re: 9.4 release notes