Re: Making tab-complete.c easier to maintain

On Wed, Dec 30, 2015 at 3:14 AM, Michael Paquier
<michael.paquier@gmail.com> wrote:
> On Sun, Dec 20, 2015 at 8:08 AM, Michael Paquier
> <michael.paquier@gmail.com> wrote:
>> On Sun, Dec 20, 2015 at 6:24 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> 2. I believe that a very large fraction of the TailMatches() rules really
>>> ought to be Matches(), ie, they should not consider matches that don't
>>> start at the start of the line.  And there's another bunch that could
>>> be Matches() if the author hadn't been unaccountably lazy about checking
>>> all words of the expected command.  If we converted as much as we could
>>> that way, it would make psql_completion faster because many inapplicable
>>> rules could be discarded after a single integer comparison on
>>> previous_words_count, and it would greatly reduce the risk of inapplicable
>>> matches.  We can't do that for rules meant to apply to DML statements,
>>> since they can be buried in WITH, EXPLAIN, etc ... but an awful lot of
>>> the DDL rules could be changed.
>
> Yep, clearly. We may gain a bit of performance by matching directly
> with an equal number of words using Matches instead of a lower bound
> with TailMatches. I have looked at this thing and hacked a patch as
> attached.

I see that you changed INSERT and DELETE (but not UPDATE) to use
MatchesN rather than TailMatchesN.  Shouldn't these stay with
TailMatchesN for the reason Tom gave above?

-- 
Thomas Munro
http://www.enterprisedb.com