README
¶
Text Streams, Redirection, and Pipelining
Every process has its own set of three text streams, collectively known as the stdio streams: stdin, stdout, and stderr. Programs can be chained together through a process known as pipelining, with the stdout of one process corresponding to the stdin of the next, and so on. Syntactically, each pair of commands in a pipeline is separated by a "pipe" character, as shown below:
# count number of files the current directory
ls | wc -l
# list first 10 error lines in a compressed log file
zcat log.gz | grep -i error | head
Stdio streams can also redirected from or to files. The following two examples are equivalent: they both lexicographically sort lines from infile, filter out duplicates, and then write the result to outfile.
sort < infile | uniq > outfile
cat infile | sort | uniq | tee outfile
Unless stdin is redirected from a file, the first process in a pipeline will receive input from the controlling terminal (the keyboard). Likewise, the stdout of the last process in a pipeline is sent to the controlling terminal, unless redirected. Even a lone command without explicit redirection is technically a pipeline, due to its implicit coupling to the terminal.
Stderr is used to communicate information not part of the normal program output, such as errors; since stderr is not normally redirected, the error output for all processes in a pipeline are displayed in the terminal.
However, it's often desirable to ignore stderr:
rm somefile 2>/dev/null
In this case, 2> means redirect stream #2 (stdin is 0, stdout is 1, and
stderr is 2) to /dev/null, in effect discarding the error output.
Example
All material is licensed under the Apache License Version 2.0, January 2004.
Directories