commandliners

Rafacas has already mentioned it, but xargs is sometimes much more useful than what it looks like. Two examples come to mind:

• Way too many files for rm or ls. It may well happen that a script has generated more than 10000 files in the same directory (it was your friend, not you, I know). If you try and rm * in there, you will be in trouble. Ditto if you simply want to count them with ls | wc -l. However, the following works:

  $ find . -type f -print0 | xargs -0 ls | wc -l
  

  (or rm instead of ls). It works because there is no globbing taking place. The -print0 option forces find to end lines using a zero \000 code, so that spaces are not taken into account as ‘field separators’. The -0 option of xargs is used exactly in this context: it sets 0 as the field separator to expect. Actually, if your friend produced normal filenames (without spaces or tabs inside), both options are unnecessary. Like in the next example.

• You want to use either the contents of a file or the result of a (probably long) pipeline to do a simple task. Say files is a list of files to be just created. Then

  $ cat files | xargs touch
  

  does the job.

There is a way to include trailing arguments (stdin is usually appended at the end of the code after xargs): use the -J command-line option:

$ ls -1d [A-Z]* | xargs -J % cp -rp % /temp/capitals

The above line will copy all the files and directories (at the present directory) starting with a capital letter to /temp/capitals. The -J option is followed by a string (in the example ‘%’) whose first (and only first) appearance will be replaced by the corresponding item from the pipe. In the example, if the files and directories in question are just AFILE.txt and A_dir, then the line above is equivalent to the two lines which follow:

$ cp -rp AFILE.txt /temp/capitals
$ cp -rp A_dir /temp/capitals

As usual, man xargs is your friend. By the way, I’d thank anyone who is able to explain the -I option to me.