How to check if the file is a binary file and read all the files which are not?

How can I know if a file is a binary file?

For example, compiled c file.

I want to read all files from some directory, but I want ignore binary files.

Use utility file, sample usage:

 $ file /bin/bash
 /bin/bash: Mach-O universal binary with 2 architectures
 /bin/bash (for architecture x86_64):	Mach-O 64-bit executable x86_64
 /bin/bash (for architecture i386):	Mach-O executable i386

 $ file /etc/passwd
 /etc/passwd: ASCII English text

 $ file code.c
 code.c: ASCII c program text

file manual page

Adapted from excluding binary file

find . -exec file {} \; | grep text | cut -d: -f1

I use

! grep -qI . $path

Only drawback I can see is that it will consider an empty file binary but then again, who decides if that is wrong?

BSD grep

Here is a simple solution to check for a single file using BSD grep (on macOS/Unix):

grep -q "\x00" file && echo Binary || echo Text

which basically checks if file consist NUL character.

Using this method, to read all non-binary files recursively using find utility you can do:

find . -type f -exec sh -c 'grep -q "\x00" {} || cat {}' ";"

Or even simpler using just grep:

grep -rv "\x00" .

For just current folder, use:

grep -v "\x00" *

Unfortunately the above examples won't work for GNU grep, however there is a workaround.

GNU grep

Since GNU grep is ignoring NULL characters, it's possible to check for other non-ASCII characters like:

$ grep -P "[^\x00-\x7F]" file && echo Binary || echo Text

^{Note: It won't work for files containing only NULL characters.}

perl -E 'exit((-B $ARGV[0])?0:1);' file-to-test

Could be used to check whenever "file-to-test" is binary. The above command will exit wit code 0 on binary files, otherwise the exit code would be 1.

The reverse check for text file can look like the following command:

perl -E 'exit((-T $ARGV[0])?0:1);' file-to-test

Likewise the above command will exit with status 0 if the "file-to-test" is text (not binary).

Read more about the -B and -T checks using command perldoc -f -X.

Use Perl’s built-in -T file test operator, preferably after ascertaining that it is a plain file using the -f file test operator:

$ perl -le 'for (@ARGV) { print if -f && -T }' \
    getwinsz.c a.out /etc/termcap /bin /bin/cat \
    /dev/tty /usr/share/zoneinfo/UTC /etc/motd
getwinsz.c
/etc/termcap
/etc/motd

Here’s the complement of that set:

$ perl -le 'for (@ARGV) { print unless -f && -T }' \
    getwinsz.c a.out /etc/termcap /bin /bin/cat \
    /dev/tty /usr/share/zoneinfo/UTC /etc/motd
a.out
/bin
/bin/cat
/dev/tty
/usr/share/zoneinfo/UTC

cat+grep

Assuming binary means the file containing NULL characters, this shell command can help:

(cat -v file.bin | grep -q "\^@") && echo Binary || echo Text

or:

grep -q "\^@" <(cat -v file.bin) && echo Binary

This is the workaround for grep -q "\x00", which works for BSD grep, but not for GNU version.

Basically -v for cat converts all non-printing characters so they are visible in form of control characters, for example:

$ printf "\x00\x00" | hexdump -C
00000000  00 00                                             |..|
$ printf "\x00\x00" | cat -v
^@^@
$ printf "\x00\x00" | cat -v | hexdump -C
00000000  5e 40 5e 40                                       |^@^@|

where ^@ characters represent NULL character. So once these control characters are found, we assume the file is binary.

---

The disadvantage of above method is that it could generate false positives when characters are not representing control characters. For example:

$ printf "\x00\x00^@^@" | cat -v | hexdump -C
00000000  5e 40 5e 40 5e 40 5e 40                           |^@^@^@^@|

See also: How do I grep for all non-ASCII characters.

Going off Bach's suggestion, I think --mime-encoding is the best flag to get something reliable from file.

file --mime-encoding [FILES ...] | grep -v '\bbinary$'

will print the files that file believes have a non-binary encoding. You can pipe this output through cut -d: -f1 to trim the : encoding if you just want the filenames.

---

Caveat: as @yugr reports below .doc files report an encoding of application/mswordbinary. This looks to me like a bug - the mime type is erroneously being concatenated with the encoding.

$ for flag in --mime --mime-type --mime-encoding; do
    echo "$flag"
    file "$flag" /tmp/example.{doc{,x},png,txt}
  done
--mime
/tmp/example.doc:  application/msword; charset=binary
/tmp/example.docx: application/vnd.openxmlformats-officedocument.wordprocessingml.document; charset=binary
/tmp/example.png:  image/png; charset=binary
/tmp/example.txt:  text/plain; charset=us-ascii
--mime-type
/tmp/example.doc:  application/msword
/tmp/example.docx: application/vnd.openxmlformats-officedocument.wordprocessingml.document
/tmp/example.png:  image/png
/tmp/example.txt:  text/plain
--mime-encoding
/tmp/example.doc:  application/mswordbinary
/tmp/example.docx: binary
/tmp/example.png:  binary
/tmp/example.txt:  us-ascii

Try the following command-line:

file "$FILE" | grep -vq 'ASCII' && echo "$FILE is binary"

It's kind of brute force to exclude binary files with tr -d "[[:print:]\n\t]" < file | wc -c, but it is no heuristic guesswork either.

find . -type f -maxdepth 1 -exec /bin/sh -c '
   for file in "$@"; do
      if [ $(LC_ALL=C LANG=C tr -d "[[:print:]\n\t]" < "$file" | wc -c) -gt 0 ]; then
         echo "${file} is no ASCII text file (UNIX)"
      else
         echo "${file} is ASCII text file (UNIX)"
      fi
   done
' _ '{}' +

The following brute-force approach using grep -a -m 1 $'[^[:print:]\t]' file seems quite a bit faster, though.

find . -type f -maxdepth 1 -exec /bin/sh -c '
   tab="$(printf "\t")"
   for file in "$@"; do
      if LC_ALL=C LANG=C grep -a -m 1 "[^[:print:]${tab}]" "$file" 1>/dev/null 2>&1; then
         echo "${file} is no ASCII text file (UNIX)"
      else
         echo "${file} is ASCII text file (UNIX)"
      fi
   done
' _ '{}' + 

You can do this also by leveraging the diff command. Check this answer:

https://unix.stackexchange.com/questions/275516/is-there-a-convenient-way-to-classify-files-as-binary-or-text#answer-402870

grep

Assuming binary means file containing non-printable characters (excluding blank characters such as spaces, tabs or new line characters), this may work (both BSD and GNU):

$ grep '[^[:print:][:blank:]]' file && echo Binary || echo Text

^{Note: GNU grep will report file containing only NULL characters as text, but it would work correctly on BSD version.}

^{For more examples, see: How do I grep for all non-ASCII characters.}

Perhaps this would suffice ..

if ! file /path/to/file | grep -iq ASCII ; then
    echo "Binary"
fi

if file /path/to/file | grep -iq ASCII ; then
    echo "Text file"
fi