Loading
Remove a path from your PATH variable
If you need to remove a path from the PATH variable before your script runs add this to the beginning of the script:
PATH=$(echo $PATH | sed -e 's;:\?/home/user/bin;;' -e 's;/home/user/bin:\?;;')
If you need, you can re-add it at the front of the list with:
PATH=/home/user/bin:$PATHOr you can re-add it at the end of the list with:
PATH=$PATH:/home/user/bin_______________________________
Related Articles
- Shell Functions and Path Variables by Stephen Collyer
- Distributed Compiling with distcc by Jes Hall
- Configuring Bash by David Blackman
______________________
Mitch Frazier is an Associate Editor for Linux Journal.
Comments
Remove a path from your PATH variable (the tr/grep version)
I've always been fond of the following structure:
RPATH="/home/user/bin"
PATH=$( echo ${PATH} | tr -s ":" "\n" | grep -vwE "(${RPATH})" | tr -s "\n" ":" | sed "s/:$//" )
Split path on ":", one per line. Squeeze double occurences of ":".
Remove line(s) that exactly matches path(s) to remove
Join list of paths to a new PATH. Squeeze double occurences of "\n".
Remove ":" at end of line.
If you need to remove more than one path, You just add to the RPATH variable
RPATH="/home/user/bin|/usr/games"
Why squeeze? If You have an empty path element, then current directory seems to be included. This can lead to call of unexpected programs.
shell replacement
you can also try the variable replacement that bash have:
echo $PATH
/sbin:/bin:/usr/sbin:/usr/bin
echo ${PATH/\/usr\/sbin:}
/sbin:/bin:/usr/bin
that is this works like ${variable/text-to-remove-in-this-variable}
so $PATH=${PATH/\/usr\/sbin:}
you can also do a find and replace:
echo ${PATH/\/usr\/sbin:/\/usr\/local\/sbin:}
/sbin:/bin:/usr/local/sbin:/usr/bin
the nasty part here is the need to escape the / character, but works well :)
higuita
Doesn't always work
What about the case where the path you want to remove is at the end? Including the colon in the pattern misses that case.
Mitch Frazier is an Associate Editor for Linux Journal.
Editing PATH variables.
A much easier way to go is to let the builtin readline functions do more of the work.
Add this to your .inputrc
"\C-xp": "PATH=${PATH}\e\C-e\C-a\ef\C-f"
What it says in english is: "If I hit ^Xp, then on the commandline say PATH=$PATH. Then use escape-control-e
to cause any variables to be expanded. Then go to the beginning of the line, then go forwards by one word and then go forward by one character."
This will leave your cursor right on the first character of the value of your PATH.
After you add your sequence to the .inputrc, just say ^X^R to cause your .inputrc to be re-read.
Steven W. Orr
If you say so :)
I'll take your word for that one. Although, I was thinking more in terms of having this in a script to take out paths you don't want, and not in terms of doing it interactively at the command line.
Mitch Frazier is an Associate Editor for Linux Journal.
That doesn't always work
Testing the above with
PATH=/home/user/bin:/one:/home/user/bin:/two:/home/user/bin
I'm getting the following result:
:/one:/two:/home/user/bin
Instead I would suggest the following
PATH=$(echo $PATH | sed -e 's;\(^/home/user/bin:\|:/home/user/bin$\|:/home/user/bin\(:\)\);\2;g')
that gives us the following result:
/one:/two
I know my test PATH is a little extreme, as you shouldn't normally see the same path more than once in your PATH. But real life systems aren't always "normal".
Not extreme at all
Actually, that's not all that extreme of an example. Yours is better.
Mitch Frazier is an Associate Editor for Linux Journal.