Wednesday 14 January 2009

Pipes

One of the original philosophies behind Unix is that each program should do one thing only, and do it well, and by and large most command-line applications in Ubuntu continue that tradition. By sticking to just that one task, they remain efficient and powerful.

But sometimes you may need more complex functionality, and this is where piping comes into its own. Piping allows you to combine commands in a simple but effective way that, if well used, makes them more powerful than the sum of their parts.

The concept is simple. Say it's a hot day and you want to water the plants in your garden. You have a tap on the outside wall of your house, but that's nowhere near the plants. You need to take the output of the tap (water) and use it as input to the plants.

How do you do it? You use a hosepipe, and that's exactly what a pipe is in Linux and Unix. It takes the output of one command and uses it as input for another.

The pipe character looks like this:
|

Where you find it differs on different keyboards, and some really do their best to hide it away! It's often shown as broken in the middle. On my Kubuntu machine it's on the same key as the backslash (\) character. Make a mental note of where it is, as you'll find it very handy!

Here's a simple example. On my Kubuntu laptop, I want to run ls -A, which shows all the files and folders in a directory, including hidden ones, within my /home directory. However, there's quite a lot here so if I display it all, it'll just run off the screen. But if I pipe it into the less command, I can easily scroll up and down through them. Here's how you do it:
ls -A | less

Try it out! This is a common way of making directories like this more readable. Note that less didn't need a file name specified, instead it accepted the output of ls -A as input. Most commands will accept this fine.

Remember I said that grep was more effective when combined with pipes? Well, here's an example where I'm again running ls -A in my /home directory. However, here I want to find files or folders with bash in their names, so I use grep to filter the output for responses that include bash:
matthew@trinity:~$ ls -A | grep bash
.bash_history
.bash_logout
.bashrc

You can use more than one pipe if you wish. To the best of my knowledge, there's no upper limit! Here's an example I used the other day when trying out OpenBSD, another open-source Unix-based operating system similar to Linux. I wasn't sure what device the CD drive was recorded as, so I ran the following:
ls /dev | grep -i cd | less

The first part listed all the files or folders in the /dev directory (in other words, devices on the computer). I then piped the output into grep, which picked out the files or folders which had cd in their names (note the use of -i to make grep case-insensitive). Finally, I piped the results of that into less so I could easily move up and down through the responses. Although this wasn't in Linux, it would work exactly the same in Ubuntu as in OpenBSD.

Pipes are extremely powerful. They let you "glue" commands together to make them more useful. You should already be able to see just how useful pipes can be if used well. Next time we'll deal with a related concept, redirection.

4 comments:

Unknown said...

my fav usage is of grep and a pipe is when you can't quiet remember the usage for a particular command but know that you used it recently:

history | grep command

I use this almost everyday at work because I have a memory like a sieve

MattBD said...

Great tip, Jon! Thanks for sharing that.
I actually found out today that you can use grep in Google searches, which is incredibly handy - for instance, enter something like ubuntu | grep icewm to search for Ubuntu and use grep to filter it for references to IceWM. This is an amazingly useful tool.

Unknown said...

I didn't know that about google, now that is a brilliant tip.

endolith said...

"I use this almost everyday at work because I have a memory like a sieve"

Why are you blaming yourself? It's not your fault. It's the fault of the command line programmers for using hard-to-remember commands, and not providing a way to find them.

Yes, you can use the "power" of the command line to compensate for its own deficiencies, but it would be better if it didn't have those deficiencies in the first place.