Someone on ##linux was talking about "bro pages", which turns out to be a site that collects random short examples of how to use Linux commands. It reminded me of Command Line Magic, a Twitter account I follow that gives sometimes entertaining or useful command-line snippets.
I hadn't been to that page on the Twitter website in a while (I usually use bitlbee for Twitter), and clicking through some of the tweets on the "Who to follow" accounts took me to someone who'd made a GNU CoreUtils cheat sheet. I didn't really want the printed cheat sheet, but I was interested in the commands used to generate it. The commands involved downloading an HTML page and didn't work any more -- the page was still there but its format has changed -- but that got me to thinking about how it might be fun to generate something that would show me a random command and its description, starting not from coreutils but from the set of all commands I have installed.
I can get a list of commands from the installed man pages in /usr/share/man -- section 1, for basic commands, and section 8, for system-admin commands. (The other sections are for things like library routines, system calls, files etc.)
So I can pick a random man page like this:
ls -1 /usr/share/man/man1/ /usr/share/man/man8 | shuf -n 1which gives me a filename like xlsfonts.1.gz.
The man pages are troff format, gzipped. You can run zcat on
them, but extracting the name and description still isn't entirely
trivial. In most cases, it comes right after the
line, so you could do something like
zcat $(ls -1 /usr/share/man/man1/* /usr/share/man/man8/* | shuf -n 1) | grep -A1 NAME | tail -1(the * for the two directories causes ls to list the full pathname, like /usr/share/man/man1/xlsfonts.1.gz, instead of just the filename, xlsfonts.1.gz).
But that doesn't work in every case: sometimes the description is more than one line, or there's a line between the NAME line and the actual description.
A better way is to use apropos (man -k), which already knows how to search through man pages and parse them to extract the command name and description. For that, you need to start with the filename (I'm going to drop those *s from the command since I don't need the full pathname any more) and get rid of everything after the first '.'.
You can do that with
it looks for everything starting with a dot (
by a digit (
[0-9] -- sed doesn't understand
followed by anything (
.*) and replaces all of it with nothing,
the empty string.
Here's the full command:
apropos $(ls -1 /usr/share/man/man1/ /usr/share/man/man8 | shuf -n 1 | sed 's_\.[0-9].*__')
Sometimes it will give more than one command: for instance,
just now, testing it, it found /usr/share/man/man8/snap.8.gz,
pared that down to just snap, and
found ten different commands. But that's unusual; most of the time
you'll just get one or two, and of course you could add another
| shuf -n 1 if want to make sure you get only one line.
man -f is a better solution: that will give a single
apropos-like description line for only the command picked by the
first shuf command.
man -f $(ls -1 /usr/share/man/man1/ /usr/share/man/man8 | shuf -n 1 | sed 's_\.[0-9].*__')
It's kind of a fun way to discover new commands you may not have heard of. I'm going to put it in my .zlogin.
[ 13:22 Oct 31, 2019 More linux/cmdline | permalink to this entry | comments ]