Unix process streams explained

I/O, or IO, stands for input/output. Anything a program does that affects or uses something outside of its own memory, like writing data to a file, or using a device (such as playing audio), is I/O.

Some terms you'll encounter a lot in the Unix world are *stdin*, *stdout* and *stderr*. These stand for standard input, output, and error, and are the main ways processes do I/O. For example, the `cat` command after reading the named file writes its contents to stdout. Normally when you run `cat` in a terminal its stdout will be the terminal, so you'll see the text of the file appear on screen. But you can make it go somewhere else instead.

The `|` character (that's a pipe, not a 1, I, or l) lets you redirect a command's stdout to be the stdin of another command. For example, if you want to use `grep` to search for text, but there are too many matches and you don't want it to flood your terminal, you can do `grep file | less`, redirecting grep's output to `less`, a command that lets you scroll through things instead of dumping them all at once.

So what's stderr?

Stdout is for the normal expected output of a command. Commands write to their stderr stream instead of stdout if they fail. For example, if you try to `grep` a file that doesn't exist, it'll print an error message to stderr, which will look the same as if it was printed to stdout. But the `|` only redirects stdout, so if you do the same `grep file | less` doesn't exist, cat won't send the error message to `sendemail` and cause it to get sent it as an email. The error message will still show up on your terminal. (It's still up to the person who wrote `sendemail` to make it not react to this by sending an empty email, but at least this way you'll find out that something went wrong.)

Other forms of stream redirection

There's a lot of other useful stuff you can do by changing where a process's stdin, stdout, and stderr are coming from or going to. For example, in most shells the `>` character is like `|` except that it writes the output to a file instead of to the stdin of another command. `dmidecode > output` would run `dmidecode` but save its output to the file named `output` instead of printing it to the terminal. You can do this with any command.

Another useful form of this is **command substitution**, which lets you use the stdout of a command as arguments to another. Most shells do this with backquotes (`). For example, "stat `cat files`" will run `cat files`, reading the text from the file named `files`, and then run `stat` on each file listed in there.

Shells do have ways of redirecting stderr, mixing it in with stdout or sending it to a different file, et cetera, but I think you get the idea now. The rest of that stuff is just a matter of looking up the syntax for your shell.


subscribe via RSS

Proxied content from gemini://yujiri.xyz/software/unix-streams.gmi

Gemini request details:

Original URL
Status code
text/gemini; lang=en
Proxied by

Be advised that no attempt was made to verify the remote SSL certificate.

What is Gemini?