ajlittoz has asked for the wisdom of the Perl Monks concerning the following question:

Hi wise monks,

I'd like to use File::MMagic to discriminate between different file types. What I basically need is text, image and "others", i.e. the first part of the MIME type from the start to the / separator. Unfortunately, File::MMagic returns invariably application/octet-stream for image file.

The first argument I gave to File::MMagic is my system "magic" file but File::MMagic result is different from system command file.

I know file is using a compiled version of the system "magic" file and I suppose both formats are equivalent (i.e. they contain the same magic numbers).

I am also sure File::MMagic method new has correctly taken note of the argument because of a bug (see File::MMagic bug?).

This behaviour has already been reported in a different context (see Comparing different MIME type checking modules). I'm not asking here for the "best" MIME type, but only for a coarse indication of the kind of data in a file. My grief is that application/octet-stream is way too coarse and that I could get a finer approximation allowed by the system "magic" file.

It looks to me File::MMagic reverts to its internal table when methods checktype_* are invoked.

1. How can I assert File::MMagic is really using the designated magic file?

2. What is a simple reliable way of classifying files as text, image and "others"?

Note: I don't want to use `file thefile_to_test` as it involves process forking, pipe connection, etc. which would negatively impact performance.

Replies are listed 'Best First'.
Re: File::MMagic returns application/octet-stream for most files
by Your Mother (Archbishop) on Jan 23, 2016 at 17:31 UTC

    Sounds like your system magic file is actually not being read maybe. Have you tried the default File::MMagic->new() just to see? Works well for me (using 1.27 on Perl 5.8.8 to do a couple tests right now) in plain cases like you describe. Can you show a minimal code example that fails for you?

[reply]

      Yes, I forced File::MMagic->new() and printed what was returned from checktype_contents(). It is invariably application/octet-stream which makes me think that the reference to the system magic file has been lost between the time I constructed a new MMagic object and it is used (although I dumped the object variable and that shows the same address).

      My version is 1.27 like yours, Perl is v5.14.4.

      Here is part of my initialisation code. As you can see, I have commented out references to magic files (the system file name is passed through hash member 'magicmime' while a fall-back file exists in the private library lib. The initial part of the file to test is read-in separately before calling one of the two functions defined in the "configuration" hash.

          my $magic = File::MMagic->new   ( # -f $self->{'magicmime'}
#                                   ? $self->{'magicmime'}
#                                   : -f 'lib/magic.mime' ? 'lib/magic
+.mime' : ()
                                    );
    $self->{'&discard'} = sub { 'text/' ne substr($magic->checktype_co
+ntents(@_[0]), 0, 5) };
#   Same, to return the complete MIME type
    $self->{'&mimetype'} = sub {$magic->checktype_contents(@_[0])};

      [download]

      Function &discard seems to work fine (at least it returns true on text files). Function &mimetype returns either text/plain or application/octet-stream on my test files.

      I may have made a big mistake in my code as I'm using a non-scope variable binding for $magic: when the functions are called, the block containing my $magic has disappeared for a long time.

[reply]
[d/l]
[select]

Back to Seekers of Perl Wisdom