file command replacement (showing the type of a file with perl)

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

As everybody surely knows, the "file" command in bash prints info about the type of a file. (i.e file flower.jpg will show that this is a jpeg image)

I'm trying to achieve the same in a perl script. Thus I wrote something like this:

use strict;
use File::Type;

my $infile = "flower.jpg";
   my $ft = File::Type->new();
     my $type1 = $ft->checktype_filename($infile);
     my $type2 = $ft->mime_type($infile);

     print "$type1\n";
     print "$type2\n";
     print `file $infile`;
[download]

and my proto-script returns

image/jpeg
image/jpeg
flower.jpg: JPEG image data, JFIF standard 1.01
[download]

Thus we have a clear winner, the File::Type module, although not bad, is clearly less accurate that file. And that's the question: ¿Do you know any way/module/function integrated with perl to obtain the same amount of info about a file? or I need to use system here to call directly the file command?. The big nuisance here is that if you search for file + perl in google you obtain millions of unrelated results

Comment on file command replacement (showing the type of a file with perl) Select or Download Code

Replies are listed 'Best First'.
Re: file command replacement (showing the type of a file with perl) by halfcountplus (Hermit) on Apr 17, 2012 at 12:41 UTC
the File::Type module, although not bad, is clearly less accurate that file. You've misunderstood something. The explicit purpose of File::Type is "to determine the MIME type of that file". 'image/jpeg' is a valid MIME type. 'JPEG image data, JFIF standard 1.01' is not. MIME types are important, eg, in internet transfers -- your browser decides how to treat something based on the MIME type in the http header. The file command (which will give you a mime type too, if you use the --mime switch) can provide more detailed information in a "human readable" form but there is no protocol for this. It looks to me like someone has written a module porting "file" into perl tho: file. However, since the author admits "this implementation is significantly slower than the C version" you might be better off just using backticks ;).	[reply]
Re: file command replacement (showing the type of a file with perl) by Corion (Patriarch) on Apr 17, 2012 at 12:44 UTC
Usually, the `file` command relies on the "magic" bytes at the start of a file. Usually these signatures are stored in a file named `/etc/magic`. There are modules that either include such a database or can reuse such a file. Most are found by Magic (resp. MMagic for File::MMagic).	[reply] [d/l] [select]
Re: file command replacement (showing the type of a file with perl) by zwon (Abbot) on Apr 17, 2012 at 13:26 UTC
If you want same output as from file have a look at File::LibMagic, it uses the same library: `$ file 1x1.png 1x1.png: PNG image data, 1 x 1, 8-bit/color RGBA, non-interlaced $ perl -MFile::LibMagic -E'say File::LibMagic->new->describe_filename( +"1x1.png")' PNG image data, 1 x 1, 8-bit/color RGBA, non-interlaced` [download]	[reply] [d/l]
Re: file command replacement (showing the type of a file with perl) by pvaldes (Chaplain) on Apr 17, 2012 at 14:04 UTC
Thank you very much for the replies, specially to halfcountplus for educate me and to Corion for point me (first) to the right trace I had tried File::MMagic but without success, it seems that is deprecated and some buggy. This is what you can read about the module in cpan: "The currently recommended module in this area seems to be File::LibMagic. Other alternatives include File::Type (gives less useful results)... Thus following the trace I finish to arrive to the same module recommmended by Zwon, (thanks also) that works like a charm # apt-get install libfile-libmagic-perl # (run this first once if you are in Debian) use File::LibMagic ':easy'; my $infile = $ARGV[0] print MagicFile($infile),"\n"; print `file $infile`; [download] Last two lines show basically the same info when we provide a filename as first argument. Problem solved thus. Thanks again	[reply] [d/l]