bug#1224: [PATCH] user/group name completion

bug#1224: [PATCH] user/group name completion

[Glenn Morris]

Version: 24.2

I am sorry for the big delay. This is now installed in the Emacs trunk.
Thank you!

bug#1224: [PATCH] user/group name completion

[Chong Yidong]

> Hello,
>
> this patch introduces a user/group name completion for
> 'call-interactively'. This feature may be used, for example, within
> 'dired-do-chown' and 'dired-do-chgrp' (although it should be extended
> to MS and MacOS systems first, BTW).

Hi Antipov, and sorry for the late response.

Thanks for writing this patch, but I'm afraid I have to turn it down.  I
don't think user/group names are important enough for `interactive' code
letters, which are a scarce resource.  You can handle them in Elisp,
instead, with an S-expression argument to interactive.

The other part of your patch, adding functions to the C code that call
getpwent and getgrent, may be useful.  But I don't see the advantage
over parsing /etc/passwd or /etc/group manually (or calling a shell
command that does it).

bug#1224: [PATCH] user/group name completion

[Eli Zaretskii]

> From: Chong Yidong <cyd@stupidchicken.com>
> Date: Sun, 12 Jul 2009 15:17:25 -0400
> Cc: 1224@emacsbugs.donarmstrong.com
> 
> The other part of your patch, adding functions to the C code that call
> getpwent and getgrent, may be useful.  But I don't see the advantage
> over parsing /etc/passwd or /etc/group manually (or calling a shell
> command that does it).

The advantage is that getpwent and getgrent are more portable, of
course.

Only Posix systems can be trusted to actually have /etc/passwd and
/etc/groups.  On systems where these files are absent, we can still
have the C APIs (or craft our owns inside Emacs, as w32.c does).

That is not to say we should necessarily add these APIs, if they
currently don't have any callers within Emacs.

bug#1224: [PATCH] user/group name completion

[Sven Joachim]

On 2009-07-12 21:54 +0200, Eli Zaretskii wrote:

>> From: Chong Yidong <cyd@stupidchicken.com>
>> Date: Sun, 12 Jul 2009 15:17:25 -0400
>> Cc: 1224@emacsbugs.donarmstrong.com
>> 
>> The other part of your patch, adding functions to the C code that call
>> getpwent and getgrent, may be useful.  But I don't see the advantage
>> over parsing /etc/passwd or /etc/group manually (or calling a shell
>> command that does it).
>
> The advantage is that getpwent and getgrent are more portable, of
> course.
>
> Only Posix systems can be trusted to actually have /etc/passwd and
> /etc/groups.

And even on those systems these files do not necessarily hold the user
database, there is NIS, LDAP,...

> On systems where these files are absent, we can still
> have the C APIs (or craft our owns inside Emacs, as w32.c does).

Please use them everywhere, manually parsing /etc/passwd and /etc/group
is broken behavior.

Sven

bug#1224: [PATCH] user/group name completion

[Juri Linkov]

> The other part of your patch, adding functions to the C code that call
> getpwent and getgrent, may be useful.  But I don't see the advantage
> over parsing /etc/passwd or /etc/group manually (or calling a shell
> command that does it).

The other part of the patch is useful for completions of "O" and "G"
in Dired.  If you decide to install the other part of the patch
please use better function names.  Since most user-related functions
in Emacs begin with the "user-" prefix (e.g. `user-real-login-name',
`user-original-login-name'), I think a better name would be `user-list'
for users, and `group-list' for groups.  These names are also consistent
with `buffer-list', `process-list', etc.

-- 
Juri Linkov
http://www.jurta.org/emacs/

bug#1224: [PATCH] user/group name completion

[Stefan Monnier]

> But I don't see the advantage over parsing /etc/passwd or /etc/group
> manually (or calling a shell command that does it).

It's always good practice to use an API rather than reimplement it.
It also will make it work right with NIS, ...


        Stefan