Is there a way to generalize environment scripts during CI/CD?

[jesshart]

NOTE: This question is a general how-to with hatch since its about an application CI/CD workflow using hatch (as opposed to a package or library workflow).

My team has adopted hatch for managing our python projects.

In the past we used Makefiles to allow alias-type one-line commands for running scripts in CI/CD, but now that we are utilizing hatch's script capabilities in the pypoject.toml we are wondering if there is a recommended approach for accomplishing something similar. We are trying to avoid having to install hatch in our CI/CD workflow since it depends on changing a DevOps resource at our company.

More generally, is there a recommended workflow with hatch and CI/CD that someone could share?

[ofek]

Sorry I don't quite understand. Are you saying you want to run scripts that are defined in pyproject.toml without using Hatch?

[jesshart]

It was a poorly framed question. Sorry about that!

We have a new hire who used Makefiles to run scripts during CI/CD. I was trying to reduce redundant scripts between local testing and CI/CD by generalizing the scripts from the pyproject.toml to the CI/CD pipeline. For example,hatch run <script_name> where the script can still be ran but without hatch being a system level requirement before the CI/CD runs.

What am I realizing about my question is that the hatch run command presupposes the environment is ready for a command.

The run command allows you to execute commands in an environment as if you had already entered it.¹

In the case of CI/CD, the environment is a combination of the run machine where requirements are specified by whatever requirements file format we provide. But we may still want to run a script defined in the pyrpoject.toml within that enviornment. For example,

# Locally
hatch run <script_name>

# CI/CD pipeline
./ <script_name>

I suppose one way to reduce redundant / wet code here would be to define a shell script in script.sh that hatch can run like this,

# pyrpoject.toml
[tools.hatch.envs.default.scripts]
app = ["./script.sh"]

# Locally
hatch run app

# CI/CD pipeline
./script.sh

Example script,

#!/bin/bash

# A simple script to print "Hello, World!" to the console
echo "Hello, World!"

In other words, maybe I answered my own question but if you have any suggestions I would love to hear them!

References

1. https://hatch.pypa.io/latest/environment/#command-execution

[ofek]

That sounds like an interesting constraint. I don't know if this works for you but there is an undocumented feature where you can force running a command without a target environment by using an empty string e.g. hatch run :foo bar. When this happens it doesn't look for predefined scripts in the normal spot but rather in a top-level [tool.hatch.scripts].