To: Bruce Momjian <bruce@momjian.us>
cc: Alvaro Herrera <alvherre@commandprompt.com>,
	Susanne Ebrecht <susanne@2ndquadrant.com>,
	Robert Haas <robertmhaas@gmail.com>,
	Grzegorz Szpetkowski <gszpetkowski@gmail.com>,
	pgsql-docs <pgsql-docs@postgresql.org>
Subject: Re: createuser/dropuser username 
In-reply-to: <201109081624.p88GO9x27859@momjian.us> 
References: <201109081624.p88GO9x27859@momjian.us>
Comments: In-reply-to Bruce Momjian <bruce@momjian.us>
	message dated "Thu, 08 Sep 2011 12:24:09 -0400"
Date: Thu, 08 Sep 2011 14:37:44 -0400
Message-ID: <26024.1315507064@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Bruce Momjian <bruce@momjian.us> writes:
> Alvaro Herrera wrote:
>> Excerpts from Bruce Momjian's message of jue sep 08 11:52:04 -0300 2011:
>>> I started going in that direction and stopped because of this example
>>> in our docs:
>>> vacuumdb --analyze --verbose --table 'foo(bar)' xyzzy

>> Maybe do it unless there are parens.

> OK, let me give that a try.

I've been thinking more about this, and realized that there is one
killer argument in favor of not auto-quoting, at least for arguments
that represent table names: if we quote, there is no way to specify
a schema-qualified name.

I'm too lazy to troll the archives to verify this, but I am moderately
sure that this point never came up in the original discussions where
we concluded that auto-quoting command-line arguments was a good policy,
because that was before we'd implemented schemas.

So what I'm thinking right now is that we indeed should not auto-quote
table- or index-name arguments.  However, usernames and database names
taken from the command line are still going to be auto-quoted, because
that's how things work in connection requests.  Is it going to be
confusing that some things are quoted and some not?

(Of course, any such change would be non-backwards-compatible and would
have to be suitably documented and release-noted.)

The backwards-compatible alternative would be to invent additional
schema option arguments so that you could write something like
"reindexdb --schema foo --table bar", but that seems pretty darn ugly.

			regards, tom lane