Re: [HACKERS] IF (NOT) EXISTS in psql-completion

Thank you for the comment.

At Mon, 6 Feb 2017 17:10:43 +0100, Pavel Stehule <pavel.stehule@gmail.com> wrote in <CAFj8pRD85cnxEEgLtOoqL-Bda2XpzvHB3a6Mr+bvf+OKpiq3Eg@mail.gmail.com>

> > 0001-Refactoring-tab-complete-to-make-psql_completion-cod.patch
> >   - Just a refactoring of psql_completion
> >
> > 0002-Make-keywords-case-follow-to-input.patch
> >   - The letter case of additional suggestions for
> >     COMPLETION_WITH_XX follows input.
> >
> > 0003-Introduce-word-shift-and-removal-feature-to-psql-com.patch
> >   - A feature to ignore preceding words. And a feature to remove
> >     intermediate words.
> >
> > 0004-Add-README-for-tab-completion.patch
> >   - README
> >
> > 0005-Make-SET-RESET-SHOW-varialble-follow-input-letter-ca.patch
> > 0006-Allow-complete-schema-elements-in-more-natural-way.patch
> > 0007-Allow-CREATE-RULE-to-use-command-completion-recursiv.patch
> > 0008-Allow-completing-the-body-of-EXPLAIN.patch
> > 0009-Simpilfy-ALTER-TABLE-ALTER-COLUMN-completion.patch
> > 0010-Simplify-completion-for-CLUSTER-VERBOSE.patch
> > 0011-Simplify-completion-for-COPY.patch
> > 0012-Simplify-completion-for-CREATE-INDEX.patch
> > 0013-Simplify-completion-for-CREATE-SEQUENCE.patch
> > 0014-Simplify-completion-for-DROP-INDEX.patch
> > 0015-Add-CURRENT_USER-to-some-completions-of-role.patch
> > 0016-Refactor-completion-for-ALTER-DEFAULT-PRIVILEGES.patch
> > 0017-Add-suggestions-of-IF-NOT-EXISTS.patch
> >   - A kind of sample refctoring (or augmenting) suggestion code
> >     based on the new infrastructure.
> >
> > 0018-Debug-output-of-psql-completion.patch
> >   - Debug logging for psql_completion (described above)
> >
> > 0019-Add-suggestion-of-OR-REPLACE.patch
> >   - Suggestion of CREATE OR REPLACE.
> >
> >
> > # I hear the footsteps of another conflict..
> >
> The patch 0018 was not be applied.

The fear came true. fd6cd69 conflicts with it but on a
comment. The attached patch set applies on top of the current
master head (ae0e550).

> Few other notes from testing - probably these notes should not be related
> to your patch set
>
> 1. When we have set of keywords, then the upper or lower chars should to
> follow previous keyword. Is it possible? It should to have impact only on
> keywords.

It sounds reasonable, more flexible than "upper"/"lower" of
COMP_KEYWORD_CASE. The additional 20th(!) patch does that. It
adds a new value 'follow-first' to COMP_KEYWORD_CASE. All
keywords in a command line will be in the case of the first
letter of the first word. ("CREATE" in the following case. I
think it is enogh for the purpose.)

postgres=# \set COMP_KEYWORD_CASE follow-first
postgres=# CREATE in<tab>
       =># CREATE INDEX hoge <tab>
       =># CREATE INDEX hoge ON emp<tab>
       =># CREATE INDEX hoge ON employee ..
postgres=# create IN<tab>
       =># create index

Typing tab at the first in a command line shows all available
keywords in upper case.

> 2. the list of possible functions after EXECUTE PROCEDURE in CREATE TRIGGER
> statement should be reduced to trigger returns function only.

Actually Query_for_list_of_trigger_functions returns too many
candidates. The suggested restriction reduces them to a
reasonable number. The 21th patch does that.

> CREATE OR REPLACE FUNCTIONS works great, thank you!

Thanks. It was easier than expected.

As the result, 21 paches are attached to this message. 1 - 19th
are described above and others are described below.

0020-New-COMP_KEYWORD_CASE-mode-follow-first.patch

  - Add new COMP_KEYWORD_CASE mode "follow-first". The completion
    works with the case of the first word. (This doesn't rely on
    this patchset but works in more cases with 0002)

0021-Suggest-only-trigger-functions-for-CREAET-TRIGGER.EX.patch
  - Restrict suggestion for the syntax to ones acutually usable
    there. (This relies on none of this patchset, though..)

regards,

--
Kyotaro Horiguchi
NTT Open Source Software Center