Last Updated: July 31, 2017
· jasny

Pipes and streams in Linux

Using stdout and stdin we can connect programs and write to arbitrary files. This level of decoupling makes it
simple to change part of the process and even allows programs written in different programming languages to be used
within a single process.

It's important not to use stdout and stdin for other types of output like progress and errors. Write that to

Output redirection

Write to file

The > redirection operator writes anything that is written to stdout to a file.

./fetch > rawdata.json

Read from file

The < redirection operator will read a file and stream it to stdin.

./process < rawdata.json

Combining read and write

You can combine the > and < operator, but you can't use it to modify a file.

./process < rawdata.json > data.json


Stream by having fetch is write to stdout and process is read from stdin.

./fetch | ./process

You can redirect the output of process to a file

./fetch | ./process > data.json

or pipe it to another process.

./fetch | ./process | ./store


The bash program tee can be used to do multiple things with the output of a process.

Pipe and store

You can specify a file to save the data to, while keeping your pipeline.

./fetch | ./process | tee data.json | ./store

Pipe to multiple processes

tee can also be used to create a subpipeline.

./fetch | ./process | tee >(./fetch-images | ./store-images) | ./store-data

Stream over the network

We can use simple TCP socket streams to pipe data to another server. This allows splitting the process across machines.

On server B ( listen to port 9000 and stream it to process and store.

nc -l 9000 | ./process | ./store

On server A we fetch and write the contents to port 9000 of server B.

./fetch > /dev/tcp/