From: Tom Lane <tgl@sss.pgh.pa.us>
To: Justin Pryzby <pryzby@telsasoft.com>
cc: Arthur Zakirov <a.zakirov@postgrespro.ru>,
        Edmund Horner <ejrh00@gmail.com>,
        PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Subject: Re: adding tab completions
In-reply-to: <14255.1536781029@sss.pgh.pa.us>
References: 
 <CAMyN-kDUKKtaZRQJu_-ahr258b1P0PC341xyH0CK3CWi7fet7g@mail.gmail.com>
 <20180529092726.GA6535@zakirov.localdomain>
 <20180603231859.GV5164@telsasoft.com> <20180604033922.GW5164@telsasoft.com>
 <20180605142941.GA25460@zakirov.localdomain>
 <20180609234212.GT10885@telsasoft.com>
 <20180611203547.GA18675@artur-book.localdomain>
 <20180627201051.GG3554@telsasoft.com>
 <20180628112037.GA6937@zakirov.localdomain>
 <20180730004243.GE3222@telsasoft.com>
 <20180730082414.GA21163@zakirov.localdomain> <14255.1536781029@sss.pgh.pa.us>
Comments: In-reply-to Tom Lane <tgl@sss.pgh.pa.us>
	message dated "Wed, 12 Sep 2018 15:37:09 -0400"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <11474.1537558154.1@sss.pgh.pa.us>
Date: Fri, 21 Sep 2018 15:29:14 -0400
Message-ID: <11475.1537558154@sss.pgh.pa.us>
Precedence: bulk

I wrote:
> The main thing that is bothering me about the remainder is its desire
> to offer single-punctuation-character completions such as "(".  I do
> not see the point of that.  You can't select a completion without
> typing at least one character, so what does it accomplish to offer
> those options, except clutter?

Actually, after poking at it for awhile, I noticed there was a much bigger
problem: tests like

    else if (HeadMatches2("VACUUM", "("))

would continue to fire even if the option list was complete, so that
after typing say

    vacuum ( verbose ) 

you would not get offered any table names, only option names.

I experimented with various fixes for that, but the only one that
worked required extending word_matches() to allow a wildcard in the
middle of a pattern.  (Before it only allowed one at the end; but
it takes just a couple more lines of code to improve that.)  With
that, we can do

    else if (HeadMatches2("VACUUM", "(*") &&
             !HeadMatches2("VACUUM", "(*)"))

and this test will not trigger if the option list is complete.

I've gone ahead and pushed it with those changes.

			regards, tom lane