Any reason to use

logging

rather than

Console

in rules.. ? Examples: https://github.com/pantsbuild/pants/blob/main/src/python/pants/core/goals/package.py#L103 https://github.com/pantsbuild/pants/blob/main/src/python/pants/core/goals/tailor.py#L495

logging

would be more appropriate for debug/trace logs which would normally never be a part of a goal’s expected output.

Ah, right. Thanks, makes sense 🙂

I think only

goal_rules

can write to the Console ?

Yeah. But even in some goal_rules, we still log things like "Wrote package to dist/..". We used to print to console, but changed The way I personally view it is to use Console when the output is the main thing the user is after, like

./pants fmt

or

./pants dependees

. Use logging when output is more of a side effect to explain the main thing, like "FYI we built this package and wrote it here"

Yeah, Console is for where the output is the goal, and say you might want to pipe to another tool.

perfect. these kind of design decisions could perhaps be added to https://www.pantsbuild.org/docs/internal-rules-architecture or similar page.. ? 🙂

Probably even better is https://www.pantsbuild.org/docs/rules-api-logging or https://www.pantsbuild.org/docs/rules-api-goal-rules#console-output-to-stdoutstderr (realized I never finished writing the first page 👀)

Right. there’s a lot of info in there, that need to re-visit. Would love to know more about streaming results, as was mentioned on that first page.. 🙂 also the effect of specifying

level

on your

@rule

decorator.. (maybe it is in there already..)

Yeah it's in part not finished because we're collectively not fully happy with the status quo With the Plugin API, I was doing "docs-driven development", where writing the docs made it clear what parts of the API were confusing/clunky etc, and then I'd fix the API as a result. I didn't have time to do that with dynamic UI yet 🤷‍♂️