Re: Tab completion issue in 9.0 beta1 - Mailing list pgsql-bugs

Excerpts from Devrim GÜNDÜZ's message of mar may 18 05:17:00 -0400 2010:
> I just noticed this:
>
> bodhi=# INSERT INTO release SAVEPOINT
>
> When there is a table called "release", even if I press V (for VALUES),
> psql replaces it with SAVEPOINT.

uh.  Yeah, you can also try "insert into commit" and it offers
completion with PREPARED, TRANSACTION and WORK.

I think those rules are lacking a check that they are anchored at the
start of the string.

This completion stuff is rather primitive.  I wonder if this could be
rewritten with a bison parser or something ... maybe even hook it on the
main parser, like the ecpg parser does.
--