Re: pg_ls_tmpdir to show directories and shared filesets (andpg_ls_*) - Mailing list pgsql-hackers

From Fabien COELHO
Subject Re: pg_ls_tmpdir to show directories and shared filesets (andpg_ls_*)
Date
Msg-id alpine.DEB.2.21.2003170828250.17584@pseudo
Whole thread Raw
In response to Re: pg_ls_tmpdir to show directories and shared filesets (andpg_ls_*)  (Justin Pryzby <pryzby@telsasoft.com>)
Responses Re: pg_ls_tmpdir to show directories and shared filesets (andpg_ls_*)  (Justin Pryzby <pryzby@telsasoft.com>)
List pgsql-hackers
About v13, seens as one patch:

Function "pg_ls_dir_metadata" documentation suggests a variable number of 
arguments with brackets, but parameters are really mandatory.

  postgres=# SELECT pg_ls_dir_metadata('.');
  ERROR:  function pg_ls_dir_metadata(unknown) does not exist
  LINE 1: SELECT pg_ls_dir_metadata('.');
                ^
  HINT:  No function matches the given name and argument types. You might need to add explicit type casts.
  postgres=# SELECT pg_ls_dir_metadata('.', true, true);
  …

The example in the documentation could be better indented. Also, ISTM that 
there are two implicit laterals (format & pg_ls_dir_recurse) that I would 
make explicit. I'd use the pcs alias explicitely. I'd use meaningful 
aliases (eg ts instead of b, …).

On reflection, I think that a boolean "isdir" column is a bad idea because 
it is not extensible. I'd propose to switch to the standard "ls" approach 
of providing the type as one character: '-' for regular, 'd' for 
directory, 'l' for link, 's' for socket, 'c' for character special…

ISTM that "lstat" is not available on windows, which suggests to call 
"stat" always, and then "lstat" on un*x and pg ports stuff on win.

I'm wondering about the restriction on directories only. Why should it not 
work on a file? Can it be easily extended to work on a simple file? If so, 
it could be just "pg_ls".

-- 
Fabien.

pgsql-hackers by date:

Previous
From: Amit Kapila
Date:
Subject: Re: Missing errcode() in ereport
Next
From: Julien Rouhaud
Date:
Subject: Re: Missing errcode() in ereport