recursing through directories

recursing through directories

[Francesco Potorti`]

I am going to add a new functionality to etags (by popular request).

Until now, when a directory name was given, etags would raise an error.
I want etags to recurse through it and look for source files.

Looking at the libc functions, I think that scandir is the most useful
for this job.  Is it standard enough?  The man page says it is
conformant to bsd4.3, so I should have no problems.  The fact is that
the etags source is very conservative about systems compatibility.

Does anyone have a snippet of code as an example of usage?  Not really
necessary, but it could help.

Re: recursing through directories

[Andreas Schwab]

Francesco Potorti` <pot@gnu.org> writes:

|> I am going to add a new functionality to etags (by popular request).
|> 
|> Until now, when a directory name was given, etags would raise an error.
|> I want etags to recurse through it and look for source files.
|> 
|> Looking at the libc functions, I think that scandir is the most useful
|> for this job.  Is it standard enough?  The man page says it is
|> conformant to bsd4.3, so I should have no problems.  The fact is that
|> the etags source is very conservative about systems compatibility.

Maybe ftw would be better, since it's POSIX (scandir isn't).

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE GmbH, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: recursing through directories

[Paul Eggert]

> From: Francesco Potorti` <pot@gnu.org>
> Date: Mon, 22 Apr 2002 16:48:32 +0200
> 
> Looking at the libc functions, I think that scandir is the most useful
> for this job.  Is it standard enough?

Not really, unfortunately.  ftw is standard, but it's a portability
hassle too.  In practice, there are real problems in running out of
file descriptors when descending into deeply nested directories.

You might try taking a look at lib/savedir.[ch] from the GNU core
utilities, written by David MacKenzie.
<ftp://alpha.gnu.org/gnu/fetish/fileutils-4.1.8.tar.gz>
It's what I use in GNU tar.  Here's a precis:

  char *name_space = savedir (dir);
  if (! name_space)
    report_error (errno, dir);
  else
    {
      char *base;
      for (base = name_space; *base; base += strlen (base) + 1)
	operate_on (string_concat (dir, "/", base));
      free (name_space);
    }

Re: recursing through directories

[Eli Zaretskii]

> From: Paul Eggert <eggert@twinsun.com>
> Date: Mon, 22 Apr 2002 11:09:30 -0700 (PDT)
> 
> > Looking at the libc functions, I think that scandir is the most useful
> > for this job.  Is it standard enough?
> 
> Not really, unfortunately.  ftw is standard, but it's a portability
> hassle too.

I agree with Paul.  I think it is best for etags to have code that
uses more portable functions such as opendir/readdir.  I suspect that
even ftw is absent on some platforms like Windows (DJGPP does have it,
though).

Re: recursing through directories

[Francesco Potorti`]

   Not really, unfortunately.  ftw is standard, but it's a portability
   hassle too.  In practice, there are real problems in running out of
   file descriptors when descending into deeply nested directories.

Do you mean that there is an open file descriptor per directory level?

   You might try taking a look at lib/savedir.[ch] from the GNU core
   utilities, written by David MacKenzie.

I'll do that.  Thanks to all for your advice.

Re: recursing through directories

[Paul Eggert]

> From: Francesco Potorti` <pot@gnu.org>
> CC: emacs-devel@gnu.org
> Organization: 
> Date: Tue, 23 Apr 2002 11:45:16 +0200
> 
>    Not really, unfortunately.  ftw is standard, but it's a portability
>    hassle too.  In practice, there are real problems in running out of
>    file descriptors when descending into deeply nested directories.
> 
> Do you mean that there is an open file descriptor per directory level?

In some implementations, yes, I think so.

It's a tricky area.  If you use an open file descriptor per directory,
you can always get back to where you came from.  If not, then you can
run into other problems.

Re: recursing through directories

[Thien-Thi Nguyen]

Paul Eggert <eggert@twinsun.com> writes:

   It's a tricky area.  If you use an open file descriptor per
   directory, you can always get back to where you came from.  If not,
   then you can run into other problems.

perhaps restricting ftw usage to "depth" mode would help.  in that case,
only a parent chain of directory descriptors is needed, i would think,
and so a smart implementation could make do w/ O(filesytstem-depth) fds
(for directories).

thi