auto-detecting a file's type

... looking only at files' contents, not at their file extension.

The relevant Wikipedia article is "content sniffing".

tools available

I really like using file(1) and magic(5) to auto-detect a file's type based on its contents. Its database covers MANY different file types, and it generally seems to get things right. However, it IS still a guess, and sometimes guesses are wrong.

Is this file text or binary?

You should be aware that auto-detecting whether a file is text is a guess/heuristic, and different heuristics often disagree with each other about whether a particular file is text.

Some tools that contain text vs binary heuristics:

• Perl's -T and -B operators [algorithm description]
  • find -type f | perl -nle 'print if -T'
• file(1)/magic(5) can guess at the file's MIME type, and MIME types conveniently mention when they're part of the 'text/*' type.
  • find -type f | perl -nle 'print if qx{file -bi $_} =~ m#^text/#'
• grep has a heuristic that allows it to ignore binary files [algorithm]
  • grep -rlI ^
  • using long argument names, that's grep --recursive --files-with-matches --binary-files=without-match ^
  • "^" means "always match, regardless of file contents"

What character encoding does this text file use?

These are always a heuristic guess, and ideally a file's character encoding should always be stated explicitly.

See also — en.wikipedia.org/wiki/Charset_detection

Some tools that can do this:

• file(1)/magic(5)
  • find -type f | perl -nle 'print if qx{file -bk $_} =~ /\bASCII text\b/'
• various Perl modules:
  • Encode::Guess and encguess — part of Perl core since v5.8.0
  • Encode::Detect
  • Encode::Detect::Detector

Is this file a Perl source-code file?

Some tools that can auto-detect if a file contains Perl source:

• based on file extensions and whether a shebang is present:
  • Perl::Metrics::Simple->is_perl_file() [algorithm]
    • find -type f | perl -MPerl::Metrics::Simple -nle 'print if Perl::Metrics::Simple->is_perl_file($_)'
  • File::Find::Rule::Perl
    • perl -MFile::Find::Rule::Perl -le 'print join "\n", find(perl_file => 1, in => ".")'
    • the above command uses the interface documented at File::Find::Rule::Procedural
• based on file(1)/magic(5):
  • find -type f | perl -nle 'print if qx{file -b $_} =~ /^perl(?! Storable)/i'
  • or, to avoid having filename characters interpreted by the shell — find -type f | perl -nle 'open(my$f,"-|","file","-b",$_)or die$!; print if <$f> =~ /^perl(?! Storable)/i'
  • regarding the above commands, when I'm looking for Perl source files, I'm usually uninterested in including binary Storable files in that list
  • Note that file often misclassifies extremely short Perl files as just ASCII.
• based on PPI, which is basically the tool that allows static analysis of Perl to happen:
  • Syntax::Check
    • (This actually malfunctions for me — it has false positives, it throws errors in places that perl -c does not, across a WIDE variety of files)
    • find -type f | xargs -n 1 syncheck
    • (the above errors out A LOT, so instead try...) find -type f | perl -nle 'system "syncheck", $_'