Showing posts with label bash. Show all posts
Showing posts with label bash. Show all posts

Saturday, December 15, 2012

Determine your BogoMIPS

Yeah, BogoMIPS are bogus, but it's a lot of fun to know how much you have in your computer. Usually you'll find this line to tell you how much you have:
dmesg | grep - 'bogomips'
(or something worse). This is a bad command line because if you happen to have an big uptime (today my uptime is 48 days, which is quite low), this command will not work as most of the kernel's logs have been flushed somewhere else. This is a bad command line because it's not fun at all!
I like dc a little bit, and here's how you can add a lot of numbers using dc:
gniourf@somewhere$ dc <<< '[+]sa[z2!>az2!>b]sb1 2 3 4 5 6 6 5 4 3 2 1lbxp'
42
gniourf@somewhere$ 
Cool eh? Please, read man dc to learn about registers and all that in dc.
To determine your BogoMIPS, you just can do this:
gniourf@somewhere$ dc <<< '[+]sa[z2!>az2!>b]sb'"$(sed -n '/bogomips/Is/.*://p' /proc/cpuinfo)"'lbxp'
10374.47
gniourf@somewhere$
If you want the average BogoMIPS per processor, do:
gniourf@somewhere$ dc <<< '2k[+]sa[z2!>az2!>b]sb'"$(sed -n '/bogomips/Is/.*://p' /proc/cpuinfo)"'zsclbxlc/p'
5187.23
gniourf@somewhere$
I wrote this note because I stumbled by accident on the most terrible command line you could think of. Just for posterity and as an example of something you should never ever do, here it is:
dumbo@moronity$ echo $(echo "print ((($( cat /proc/cpuinfo | grep bogomips | awk '{print $3}' | tr -s '\n' '+')0)/$(echo 4))+($( cat /proc/cpuinfo | grep bogomips | awk '{print $3}' | tr -s '\n' '+')0))/2" | python)
By respect to the author of this junk, I will not give you his name.


Exercise.
  1. Count how many pipes, subshells and process forks there are in the previous command line.
  2. Count how many Useless Uses of Cats there are.
  3. Conclusion? (Hint: look at the login)

Cheers!

Sunday, November 13, 2011

A new bash implementation of ed

You know ed, right? Sure you do, because ed is the standard Unix editor! Last night I was working on a computer that didn't have ed installed, so I quickly wrote a bash implementation. It's not that hard after all! Check it out:
gniourf@somewhere:~$ while ((1)); do read; echo '?'; done
All right, apart from saying stupid things in this note, I'd just like to say that ed can be really cool to format files from a bash script. Most people use sed quite fluently, and are always tempted to solve some problems with it in scripts. ed can be useful because it's the standard Unix editor a genuine text editor after all! In a script you'll use a here document. For example, if you want to change the 40th line of a file named file to hello, gniourf_gniourf you'll probably do:
ed -s file <<EOF
40c
Hello gniourf_gniourf
.
wq
EOF
The -s option so that the sucker remains silent.
If you liked ed, you'll probably also like ex!

Cheers!

Saturday, November 12, 2011

What are the groups I don't belong to?

So you're running your nice linux box, and everything is fine. Everything? No. All of a sudden you really wonder what groups you're in. No problem, just issue
gniourf@somewhere:~$ groups
and you'll be okay. Hey, but wait, in fact that was not exactly your question. In fact, after a deep thought, you realize that your question was the opposite! You were wondering what groups you're not in! If you know a little bit of sed, and if you know how the entries of the /etc/group are formatted, you realize that this one-liner will do the job!
gniourf@somewhere:~$ sed "/\b$USER\b/d;s/:.*//" /etc/group | tr \\n ' '; echo
The piped tr is just here because you don't want to have only one group per line, and the echo to have a new line after all this junk.
You could even add an alias in your .bashrc, or make a function, depending on your needs.

Cheers!

Don't use backticks in bash!

No, don't, it's very ugly. And I don't know how I would read something like this:
gniourf@somewhere:~$ echo `echo `echo hello``
You know, in more modern bash, we use $(...) instead of `...`. So what do you think of the previous line? Which of the following two possibilities is it?
gniourf@somewhere:~$ echo $(echo $(echo hello)) # Possibility 1
gniourf@somewhere:~$ echo $(echo )echo hello$() # Possibility 2
You'll figure it out by yourself, but you probably understand that the $(...) is more robust, since it doesn't leave any ambiguity whatsoever: nestings are neat!

Cheers!

Never parse the output of ls, ps, etc... Use globs or find or something else!

It's all in the title!
So, you want to delete all files that end with .back, and (because you read a terrible advanced bash guide) you're tempted to do this:
gniourf@somewhere:~$ for i in `ls *.back`; do rm $i; done
The problem is that this will miserably fail whenever a filename contains a space. (Try it!). Also, the backticks are ugly (but it's not the worse thing here). Instead, do:
gniourf@somewhere:~$ for i in *.back; do rm "$i"; done
(also observe the quotes). You should realize that bash's globs are infinitely better than parsing the output of a program. If you want to use something like this in a script, maybe you should shopt -s nullglob or shopt -s failglob. For example, try:
gniourf@somewhere:~$ echo *hello # assuming that there's no file ending with hello
The output will be something like:
*hello
Because when a glob doesn't match, it is used verbatim. If you want a non-matching glob to expand to nothing, you must set nullglob on thus:
gniourf@somewhere:~$ shopt -s nullglob
Now,
gniourf@somewhere:~$ echo *hello
outputs nothing.
If you want a non-matching glob to generate an error, you must set failglob on thus:
gniourf@somehwere:~$ shopt -s failglob
Now,
gniourf@somewhere:~$ echo *hello
bash: no match: *hello


Cheers!