Shell Scripting: Expert Recipes for Linux, Bash, and More

Pipelines

Pipelines are a central feature of Unix and Linux shells. A pipe connects two processes together, generally attaching the standard output of one process to the standard input of another. Instead of writing the output of the first command to a file and then running the second command, taking that file as input, the intermediary file can be bypassed entirely with this method. It is so central to shell scripting that it has already been mentioned in passing many times, but it is worth clarifying what actually happens when a pipe is set up. This example pipes the output of find into grep:

$ find / -print | grep hosts
/lib/security/pam_rhosts.so
/var/lib/ghostscript
/var/lib/ghostscript/CMap
/var/lib/ghostscript/fonts
/var/lib/ghostscript/fonts/cidfmap
/var/lib/ghostscript/fonts/Fontmap
find: '/var/lib/php5': Permission denied
find: '/var/lib/polkit-1': Permission denied
/var/lib/dpkg/info/denyhosts.postinst

What happens here is that grep is started first, followed by find. Once both processes exist, the output of the find process is linked to the input of the grep process. The find process can then run and its output gets sent to grep. As the preceding sample output shows, the standard error device for the find process is still the calling terminal, so that is displayed as normal. It has not been passed through grep. This is evident because the error lines do not include the text “hosts” anywhere in them. This, therefore, is the standard error of find and not the output ...