nuprompt

My new bash prompt. Replacing https://github.com/astromechza/prompt.

# Install via cargo
cargo install --git https://github.com/astromechza/nuprompt

# The following is added to the bash rc or profile:
eval $(nuprompt bash)

# This expands into PS0 and PROMPT_COMMAND. PS0 captures the previous command start time, while PROMPT_COMMAND generates a
# PS1 line with the exit code, duration, user, cwd, and git information.

PS0='$(nuprompt ps0 $$)'
PROMPT_COMMAND='eval $(nuprompt ps1 $$ $?)'

• $NUPROMPT_RUST_LOG=debug will display debug logging information on stderr
• $NUPROMPT_NO_GIT=true will disable git status detection which may increase performance if you're operating in a large git repo
• $NO_COLOR=true will disable coloured output (this happens automatically if stdin is not a terminal)

FAQ

Why write a binary? why not use the PS1 formatting characters and pure bash?

Because I want to.

Why no cargo crate and binary releases?

I'm the only intended audience. I can install or build it via cargo install --git when needed.

Why do you use bash and not another shell?

Tried ZSH in the past. Got too distracted customising it. Wanted something standard that was more likely to be installed everywhere.

Background

I'll be using the following bash things:

• PROMPT_COMMAND - a command to run before displaying the prompt via PS1. The output of this command goes to stdout just before displaying PS1.
• PS1 - is the prompt variable to render (not just print) on the left on a new prompt line. This can contain special variables and colour codes.
• PS0 - an expression to evaluate just before executing the command captured by the prompt. Any output goes to stdout just before executing the command.

So what can I do with these to make it as easy to understand and install as possible?

Let's start with what I want to represent in my prompt:

1. The current working directory with $HOME replaced by ~
2. The git status if we're in a git directory
3. The exit code of the previous command (if not 0)
4. The current username or id
5. The execution time of the previous command

Firstly, lets assume that PS1 is set by PROMPT_COMMAND in some way. Either, by having a bash function to capture the output, or by outputting some eval-able expression that sets it. So this can capture (3) and calculate (1), (2), and (4).

Then, we need to capture the time of the previous command. We can do this by using PS0 to store the start time of a command in a temporary file associated with the shell process id. The bash special parameter ${$} can be used to capture the PID and then write this to a temporary file. Then in our main prompt, we'll try to read this file and use the contents to calculate an elapsed time.

This leaves us with the form of PROMPT_COMMAND='eval $(nuprompt ps1 ${$} ${?}) (with ${?} capturing the previous exit code, and ${$} capturing the shell process id). One thing I did in the previous version of my prompt was to insert a newline if the prompt wasn't at the left hand margin. This could happen if the previous command output didn't have a newline itself, however, after using this for some time I found that it wasn't clever enough and would frequently break commands if I typed them too fast.