Save last working directory on Bash logout

问题:

Is it possible to save the last working directory when you do an "exit" in bash. So, the next time you login, it will be at the directory you were at when you last logged out.

回答1:

Save the last working directory in ~./.bash_logout into a hidden file:

pwd > ~/.lastdirectory

Read this file in ~/.bashrc with

[ -s ~/.lastdirectory ] && cd `cat ~/.lastdirectory`

回答2:

Put this in your ~/.bashrc (probably at the end):

cd "$(<~/.storepwd)"

Then you have a choice of the following:

Either put this in ~/.bash_logout (will not remember directory for interactive non-login shells):

printf %s "$PWD" > ~/.storepwd

Or use a trap on EXIT, put this in your ~/.bashrc (traps can be overwritten easily/accidently):

trap 'printf %s "$PWD" > ~/.storepwd' EXIT

Or use PROMPT_COMMAND, put this your ~/.bashrc (probably the most reliable way):

PROMPT_COMMAND+='; printf %s "$PWD" > ~/.storepwd'

Also make sure that your ~/.profile or ~/.bash_profile sources your ~/.bashrc, otherwise the path won't be restored when login shells start.

See http://mywiki.wooledge.org/DotFiles for information on how you should use your dotfiles.

回答3:

You also need to be careful with this entire concept. If, for example, you manage to get your code run when you don't expect, it can confuse any number of things.

For example, if you were to run a shell script that changed the current working directory, and then, during exit, managed to save its CWD, you may end up, in the next shell that starts up, in a directory you don't expect.

Another popular one, at least for me at $work, is when a script is run when root su's to your user to run a script, all automatically (i.e., programmatically). If you're doing this CWD trick for your own user, that's fine, but it can majorly screw up with system tools that expect that "su - youruser -c ls" returns the files in your home directory. (Ok, calling "ls" as your user is a dumb idea, but there are things where your user may have a better environment for running something than root does - e.g., on NFS mounts that have root squashed.)

So, the key here is to tell if you're in an interactive shell or not, and, if not, don't restore or save your CWD. IIRC, $- has an 'i' for interactive shells. Use it for setting the trap and for restoring the CWD.

Personally, I just set an alias:

alias go='. go.sh'

and then go.sh simply checks its parameter for which directory I want to go to. e.g., "go myproj" will cd ~/svn/myproj, and "go bigdisk" will cd /mnt/bigdisk. And, for more complex cases, it can take more parameters, e.g., "go myproj lib" would just "cd ~/svn/myproj/myproj/lib" or whatever. No automatic saving, but none of the side effects, either.

回答4:

You can write current directory (pwd) to some file every time when changing dir (cd). When starting terminal, bash startup script should look that file for "last dir". If something found - cd to it.

回答5:

Look at the .bashrc and perhaps being able to modify it at logout time to contain the directory you are interested in.

回答6:

You most likely want to save the current directory to a file when the shell exits. There are a number of ways to detect the shell exitting:

1. Put some code in ~/.bash_logout. This will only be executed when a login shell exits, so if you were to just close a terminal window, the directory would not be saved.
2. Trap on EXIT. Using the "trap" builtin, you can have some code executed when the shell exits. This will run when any shell exits - well, any shell in which you set up the trap, that is.

You could save the CWD at every prompt, or every time you run 'cd', but that's a bit overkill.

To restore your CWD when the next shell starts is a simple matter of running 'cd' on the contents of your save file if it exists.

You may want to think about what should happen if you have multiple shells running at the same time. Which one do you want to save the CWD of? Should all new shells that start use that CWD? Or only just the next shell that starts?