Date: Sun, 7 Jun 2020 10:07:19 +0200 (CEST)
From: Fabien COELHO <coelho@cri.ensmp.fr>
To: Justin Pryzby <pryzby@telsasoft.com>
cc: Alvaro Herrera <alvherre@2ndquadrant.com>,
    David Steele <david@pgmasters.net>, pgsql-hackers@postgresql.org,
    "Bossart, Nathan" <bossartn@amazon.com>,
    Thomas Munro <thomas.munro@gmail.com>, Tom Lane <tgl@sss.pgh.pa.us>
Subject: Re: pg_ls_tmpdir to show directories and shared filesets (and
 pg_ls_*)
In-Reply-To: <20200526021003.GW4472@telsasoft.com>
Message-ID: <alpine.DEB.2.22.394.2006070908120.32228@pseudo>
References: <20200316214806.GQ26184@telsasoft.com>
 <20200316221736.GA15635@alvherre.pgsql>
 <20200317031427.GU26184@telsasoft.com>
 <alpine.DEB.2.21.2003170828250.17584@pseudo>
 <20200317190401.GY26184@telsasoft.com> <20200331200812.GI14618@telsasoft.com>
 <alpine.DEB.2.21.2004121145190.5257@pseudo>
 <20200503024215.GJ28974@telsasoft.com> <20200526021003.GW4472@telsasoft.com>
User-Agent: Alpine 2.22 (DEB 394 2020-01-19)
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII; format=flowed
Precedence: bulk


Hello Justin,

> Rebased onto 7b48f1b490978a8abca61e9a9380f8de2a56f266 and renumbered OIDs.

Some feedback about v18, seen as one patch.

Patch applies cleanly & compiles. "make check" is okay.

pg_stat_file() and pg_stat_dir_files() now return a char type, as well as 
the function which call them, but the documentation does not seem to say 
that it is the case.

I must admit that I'm not a fan on the argument management of 
pg_ls_dir_metadata and pg_ls_dir_metadata_1arg and others. I understand 
that it saves a few lines though, so maybe let it be.

There is a comment in pg_ls_dir_files which talks about pg_ls_dir.

Could pg_ls_*dir functions C implementations be dropped in favor of a pure 
SQL implementation, like you did with recurse?

If so, ISTM that pg_ls_dir_files() could be significantly simplified by 
moving its filtering flag to SQL conditions on "type" and others. That 
could allow not to change the existing function output a keep the "isdir" 
column defined as "type = 'd'" where it was used previously, if someone 
complains, but still have the full capability of "ls". That would also 
allow to drop the "*_1arg" hacks. Basically I'm advocating having 1 or 2 
actual C functions, and all other variants managed at the SQL level.

-- 
Fabien.