A queue for shell commands in emacs
Note:
This is in a very early alpha. Also this project is just for fun.
- Queue arbitrary shell commands.
- Execute the queue automatically.
- Send commands from the command line
- Cancel/Kill Jobs
- Multiple queues
- Permanent build history and logs
- Select git branch for each build
- Select stashes for each build
- Build Management buffer
- Hooks can add metadata to jobs
- Richard Stallman Quotes Plugin
- Complex dependency graphs
- Translate zsh function to bash
- Metrics
- Commit hook
After you installed emaci, load it via:
(require 'emaci) (emaci/load-vars)
Use emaci/submit-job
or emaci/submit-job-comint
to queue shell commands.
The queue will get executed immediately. While the queue is running,
you can submit more jobs. Once a job had finished, it will execute the next in the queue.
You can specify a branch that will get checked out before the command is run.
The original branch will be checked out after the command has finished.
The same goes for stashes. You can apply as many stashes as you like,
which will get reverted afterwards.
Jobs are queued in emaci-queue
and archived in emaci-history
.
You can cancel or kill the currently running job
with emaci/cancel-job
or emaci/kill-job
. If there are more jobs in the queue,
you have to restart the queue by calling emaci/execute-next
.
The output, queue and history is saved to the emaci-save-dir
directory.
To load them execute (emaci/load-vars)
after setting emaci-save-dir
.
emaci-save-dir
defaults to "~/.emaci/"
.
To view the queues call emaci/mgmt-buffer
. This will show the emaci management buffer.
You can expand and hide sections under your cursor by pressing TAB
.
Pressing RET
with the cursor on a job, will show you the output.
Use c
or k
for canceling/killing a job.
By default only the last 50 jobs are shown.
You can customize this by setting emaci-max-history-len-status
.
Like any other good CI-Server emaci
has a quotes plugin.
Enable Richard Stallman quotes by setting emaci-enable-rms
to t
.
Having a queue for shell commands has a lot of use cases.
I often find myself waiting for multiple code reviews to finish. Often they get completed in bulk and I have to merge multiple feature branches. Before each merge, I want to merge the latest dev branch into my feature branch and run the tests one last time, to see if it really does work. Because I have to wait for each test before I can test the next branch, scheduling comes in handy. Also sometimes I have multiple configs/environments for one project but for various reasons (not in my power to change) can't test them in parallel.
There are already some solutions for emacs like shell-command-queue.el or command-queue.el. But they lack some functionality to use emacs as a simple build management system, when you can't (or don't want to) use a proper build/test server.
Emacs has a built-in server. To start the server call:
(server-start)
Now you can start emacsclients in any shell, and the clients will use the running server (e.g. to load faster). This allows for emaci to be used like a CI-server.
To use a convenient (zsh) shell function to send commands to emacs
source emaci.zsh
in the emaci directory or copy and paste it in your .zshrc
or
somewhere on your fpath
.
Then you can use:
$ emaci echo Hello World
To submit a job via the command line use:
$ emacsclient --eval "(emaci/submit-job-comint nil \"$PWD\" \"echo Hello World\")"
Note:
You can also use emaci/submit-job
and specify the mode of your compilation buffer yourself.
It is important that you use double-quotes so the current shell working directory is replaced in the string. You could wrap this command in a shell function to make it easier to submit commands.
Unfortunately the current environment is not used. It's only possible by setting the environment in the shell command argument. You could however automate this via a shell function.
zsh:
$ emacienv () { echo -e $(declare -px | awk '{if (NR == 1) printf $0;else if ($0 !~ /^typeset -.*/ && last !~ /^typeset -ax.*/) printf "\\n"$0;else printf " && "$0;}{last=$0}')' && ' }
This uses declare
to 'serialize' all environment variables to typeset
commands. These can be evaluated to restore the exact same environment. awk
is used to concatenate the list of commands with &&
so we get a long one-liner. The awk
command uses some logic to preserve newlines in environment variables.
Now we can use this environment replication mechanism to give us a nice command.
zsh:
$ emaci () { emacsclient --eval "$(echo "(emaci/submit-job-comint nil \"$PWD\" \"$(emacienv)cd $PWD && $@\")")" }
Note:
For some reason there is some weird behaviour with the working directory
if we don't add the cd $PWD
command.
Now we can use this to send shell commands to emacs:
$ emaci echo Hello World $ emaci "./configure && make && make install" $ emaci 'echo $PWD'
I find it somehow amusing.