Sed: Extract Patterns From Files and Texts

In this article, extracting a pattern from a text file and from another output will be demonstrated. In order to do this we will use sed though this is not the only approach to this problem. The importance of this problem is that users face with similar text manipulation very frequently especially when they write shell scripts.

Getting a Synthetic Data

Consider you run the following command, and you would like to get the card’s name.

wpctl inspect @DEFAULT_SINK@

>id 53, type PipeWire:Interface:Node
>    alsa.card = "1"
>    alsa.card_name = "SteelSeries Arctis 1 Wireless"
>    alsa.class = "generic"
>    alsa.device = "0"
>    alsa.driver_name = "snd_usb_audio"
>    ...
>  * object.path = "alsa:pcm:1:iec958:1:playback"
>  * object.serial = "11163"
>  * priority.driver = "1008"
>  * priority.session = "1008"

Turning Selected Line to Regex

This piece of text is actually the output of Wireplumber’s inspect command on my machine with some truncated parts. Now, suppose that we want to extract what alsa.card_name is equal to. First, we need to generate the regex equivalent that exact line.

    alsa\.card_name = "SteelSeries Arctis 1 Wireless"

The line starts with four spaces, we can match to this with \s*, then we have to match our variable name alsa\.card_name = ". The text in the double quotes is what we aim to get from out command. We can group this piece of information as \(.*\), here backslashes and braces group whatever inside them dot star will match any length of any character. Lastly, we will add closing quote to the pattern.

This is the regex equivalent of the line we selected.

\s*alsa\.card_name = "\(.*\)"

Extracting the Regex Line

Only thing left is to call sed command. We are passing -n to suppress printing of the non-matching lines. s/ at start will substitute our matching pattern with \1 which refers to first group. /p at the end will print the matching line for us.

Getting the Desired Pattern

The most common ways to extract patters are from a file and from another commands output. Both are the extractions with sed are very straightforward to apply.

Extract Pattern From a File

Here we directly call sed to find desired regex pattern from a file called file.txt.

sed -n 's/\s*alsa\.card_name = "\(.*\)"/\1/p' file.txt

Extract Pattern From Pipe

Below, we are extracting the output of some other command called program. In order to do this, we pipe the output to sed with | symbol.

program | sed -n 's/\s*alsa\.card_name = "\(.*\)"/\1/p'