Have you ever asked yourself "how did I run that obscure command with a bajillion arguments 5 years ago, and where?" And wished that the incantation had been preserved in your shell history?

While you can configure bash to keep a very long history, there are still some caveats, namely the lack of concept of session and the lack of working directory information. If you have more than one shell open, history can get confusing. And if you work on multiple hosts, even more so. Finally, bash history files are surprisingly easy to clobber by accident.

The Distributed Shell History Database, dshdb, keeps your history forever in an sqlite database; preserves metadata such as session, CWD, hostname, elapsed time, and exit status; makes searching the database easy; and finally, makes it easy to synchronize the databases between multiple hosts.

It is written for Python 2 and 3, since many Linux distributions still in use only have Python 2 by default. Hooks for bash are provided; I think it's not too difficult to adapt them to zsh, but I don't use zsh. If you get it working, please send it along!

First, install Bash-Preexec. This is the only dependency (not counting Python and bash!).

Then, run make install to copy the scripts to ~/.hist. You can choose a different location by adding HIST_DIR=<your_location> when running make. This directory will also be the location of the history database file, hist.db.

Finally, add a line at the bottom of your .bashrc to source hist.sh; for example, if you used the default location,

. ~/.hist/hist.sh

Optional: define a more convenient alias. I have alias hi=hist.py.

The default invocation just shows the last 30 commands in reverse chronological order, showing timestamp and command.

$ hist.py
2021-12-07T09:35:20	ls
2021-12-07T09:35:02	git log
...

You can constrain the output to a given directory or session, search for a command substring, search by exit status or elapsed time. See the help message for details. For example, let's search for make commands in the current directory with a nonzero exit status:

$ hist.py -d . -x '!= 0' make
2021-12-15T11:01:35	make test
2021-12-15T10:50:59	make
2021-12-14T10:45:46	make
...

You can also control the output format to choose which columns are included in the tab-delimited output. For example, to show directory, elapsed time, exit status, and command.

$ hist.py -fdexc
~/hist	0	1	false
~/hist	0	0	true
~/hist	3	0	sleep 3
...

To synchronize your database with your database on example.com, simply say

$ hist.py --sync example.com

Note that the syncronization goes both ways!

HISTTIMEFORMAT="%s%t" history | hist.py

Timestamps have 1-second resolution; if you run the same command twice within the same second on the same directory, hostname, shell session, with the same elapsed time and exit status, it will be inserted into the database only once.

If you use the edit-and-execute-command readline binding in bash (C-xC-e by default) to run commands in multiple lines, all of them will end up with the same metadata, other than the timestamp. For example, if your edited command is

sleep 1
sleep 2
false

all three commands will be recorded with an elapsed time of 3 seconds and exit status 1.

After I started this project, I came to realize that there were a couple of similar projects already out there: Bashhub and RASH. They may be more featureful and polished, doing things such as interactive search or uploading your history to the cloud. My approach is more minimalistic, and I don't plan to add those features. I also tried to minimize the dependencies so that the only thing you need to install is a python script and a shell script. It works exactly the way I want. This is my wheel. There are many like it, but this one is mine! :-)