Re: new VFS directory listing parser

[Roland Illig <roland illig gmx de>]

Leonard den Ottolander wrote:
> Hi Roland,
>
> On Tue, 2005-09-27 at 01:27 +0200, Roland Illig wrote:
>
> > The current directory listing parser (for ftpfs and extfs) has problems 
> > with file names starting with white-space or a four-digit sequence.
>
>
> The latter is caused by the fact that different types of time format are
> accepted: "Mon DD YYYY", "Mon DD hh:mm", "Mon DD YYYY hh:mm" and "Mon DD
> hh:mm YYYY". This makes it so the code doesn't know if it parsed the
> full date after having parsed the first 3 fields.
>
> If you've just dropped the assumption there are more than 3 fields and
> only accept "Mon DD YYYY" and "Mon DD hh:mm" as valid formats the
> existing code can quite easily be patched instead of doing a whole
> rewrite. See utilvfs.h around line 635 ("This is a special case for
> ctime"). Add a var got_time and stop parsing the date after either
> got_time or got_year.

After I had rewritten vfs_parse_filetype and vfs_parse_fileperms (see 
utilvfs.c), I found the interface of these functions general enough so 
that it could be extended to the whole parsing of directory listings. I 
think one benefit of my code is that you can use all the parsing 
functions the same way. If you have used any of them, you know how to 
use all others. This has not been the case with the old code.

My code makes it easy to add support for other file listing types, as 
every component of a listing line (filemode, size_or_rdev, filedate, 
uid, gid) has its own parsing function. If you look at the code of 
vfs_parse_unix_line(), this should become very clear.

I also didn't like the function names. I expect that a function called 
is_year(...) does not modify its parameters (see the Java Coding 
Conventions). Most functions also used sscanf() for parsing, which is 
inappropriate, as sscanf("%d", "123XXXXXXXXXXXXXXX", &n) will return 
123, although "123XXXXXXXXXXXXX" does not represent a number.

And yes, I have dropped that assumption. I have started to collect 
example listing from some FTP servers, but they are still too few. I 
will also examine the Indy FTP code (see http://www.indyproject.org/), 
which has some more examples.

Roland