how to pass pipe output to file command

nickleus · 06-19-2008, 04:23 AM

i want to recursively find all files in a folder and subfolders that are encoded as utf-8, using bash.

my verbal structure so far looks like this:
recursively find all files in folder and subfolders|file --mime filename|grep UTF-8

but i hit a wall at the "file --mime" part. i've tried testing it like this:

Code:

ls | grep index.jsp | file --mime

but i get the following error:

Quote:

Usage: file [-bcikLhnNrsvz0] [-e test] [-f namefile] [-F separator] [-m magicfiles] file...
file -C -m magicfiles
Try `file --help' for more information.

anybody have a solution?

thanks in advance =)

colucix · 06-19-2008, 04:31 AM

Piping the output of ls passes a string to the file command and it is not interpreted as a file. You can use find with -exec option or use xargs, for example

Code:

find . -name index.jsp -exec file --mime {} \;

or

Code:

ls | grep index.jsp | xargs file --mime

nickleus · 06-19-2008, 04:41 AM

weird. it passes a *filename* as a string, so i would think that would be good enough, but apparently not.
xargs seems way cleaner. i think i'll use that. i fear find, just like garth fears change.
wicked. thanks for the quick reply =)

Nick

nx5000 · 06-19-2008, 04:45 AM

or
ls | grep index.jsp | file --mime -

pwc101 · 06-19-2008, 04:50 AM

Or:

Code:

ls | grep index.jsp | file --mime -

The man page states:

Quote:

Originally Posted by man file

-f, --files-from namefile
Read the names of the files to be examined from namefile (one per
line) before the argument list. Either namefile or at least one
filename argument must be present; to test the standard input,
use '-' as a filename argument.

What this method doesn't do though, is preserve the filename section in the output of the file command. So instead of getting

Code:

index.jsp: some-file-type

you get

Code:

/dev/stdin: some-file-type

colucix's examples don't have this problem, but I thought I'd point out the '-' as a filename, meaning standard input, which can be used in many programs (tar, bzip etc.).

edit: Damn, I type too slowly!

nickleus · 06-19-2008, 05:03 AM

awesome answers nx5000 and pwc101. thanks a lot for info =) i am smarter now.

jschiwal · 06-19-2008, 06:18 AM

Code:

ls | grep index.jsp | file --mime

This looks like you want the mime type of the index.jsp file. For that you could use "file --mime index.jsp".

Unless you have files of the pattern "*index.jsp". For that you can use "file --mime *index.jsp".

If the files to test have a certain pattern, you can wildcards instead of using ls or find.

file --mime dir1/*index.jsp dir2/*index.jsp *index.jsp

---

A caveat is when you have a very large number of files in the directory matching the pattern. This could cause an out-of-memory error in the shell because you exceed the maximum number of arguments. In that case, pipe find to xargs as in colucix's second example and use the -n or -L option. For your original goal of finding the mime type of all files, you will probably want to do this.

Also, because filenames may contain white space characters, you will want to use the -print0 option to find and the -0 option for xargs:

Code:

find ./ -type f -print0 | xargs -0 -L 500 file --mime >file-mime.list