> On Oct 11, 2021, at 4:49 PM, Tom Lane <t...@sss.pgh.pa.us> wrote:
>
> You are attacking a straw man here. To use a period in an identifier,
> you have to double-quote it; that's the same in SQL or \d.
That's a strange argument. If somebody gives an invalid identifier, we
shouldn't assume they know the proper use of quotations. Somebody asking for
a.b.c.d.e is clearly in the dark about something. Maybe it's the need to quote
the "a.b" part separately from the "c.d.e" part, or maybe it's something else.
There are lots of reasonable guesses about what they meant, and for backward
compatibility reasons we define using the suffix d.e and ignoring the prefix
a.b.c as the correct answer. That's a pretty arbitrary thing to do, but it has
the advantage of being backwards compatible.
>> I expect I'll have to submit a patch restoring the old behavior, but I
>> wonder if that's the best direction to go.
>
> I do not understand why you're even questioning that. The old
> behavior had stood for a decade or two without complaints.
I find the backward compatibility argument appealing, but since we have clients
that understand the full database.schema.relation format without ignoring the
database portion, our client behavior is getting inconsistent. I'd like to
leave the door open for someday supporting server.database.schema.relation
format, too. I was just wondering when it might be time to stop being lenient
in psql and instead reject malformed identifiers.
—
Mark Dilger
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
