Workflow/rationale for multiple environments?

[doolio]

I'll preface this discussion by stating that I'm still very much a beginner programmer so please excuse my ignorance in what I state below.

Is hatch unique in creating multiple (virtual) environments by default namely default (but takes the project name), lint and all (previously called test in the docs?) which allows you to select a specific python version environment from a matrix if I'm not mistaken?

Regardless, is the rationale behind this that you only activate say the lint environment when you want to lint your project, a test environment when you want to test your project, a doc environment when you want to produce the documentation for your project etc.?

I ask as it seems the default linting tools (black, mypy and ruff) are only installed into the lint environment. But that means those tools are not available to you in the default environment, right? They don't seem to be. But wouldn't you want such tools available when working on the project?

I know one could add them as dependencies in the default environment but am interested to know why hatch doesn't by default. I'd imagine that would make it easier for beginners such as myself that are not so familiar with virtual environments. Perhaps it is done deliberately to encourage beginners to look into this layout as I have so they understand how and where dependencies will be installed.

Or perhaps the idea is that one creates lint, test environments etc. that could then be used by any project? But I don't think hatch works that way by default even if all environments are centrally located or am I mistaken?

[ofek]

It's better to not pollute the environment(s) where your project is installed with extra dependencies to more precisely match the end-user and also prevent dependency conflicts

[doolio]

Sorry, the lint environment.

[ofek]

I think not, rather whatever environment has the project

[doolio]

OK, thanks. This is what I was looking for. The best practices to follow.

[doolio]

but by default hatch installs coverage and pytest in the same environment where the project is installed.

Could you clarify the rationale for this then considering the argument in your first reply?

[ofek]

I suppose extra effort? It's tricky to have tests work when what is installed is separate from the test runner. It's possible but I haven't put much thought toward it