Hey Pants Team, curious if there's a way we can retrieve granular runtimes from the lint goal. For example, if we run `flake8`/`isort` on lint we want to know the total runtime for each lint tool

the workunits that Pants records contain this information, and so plugins can emit it. Toolchain’s plugin does, for example.

okay that is cool, so pants records this can we can access it in via a separate plugin. So we would run

./pants lint ::

and then

./pants our-workunit-inspector-plugin

So we would run

./pants lint ::

and then

./pants our-workunit-inspector-plugin

not quite… the relevant plugin interface is

StreamingWorkunitHandler

: it receives workunits during the run, rather than being a separate goal.

I found this test and I'm getting an idea how this works. Is there any other examples of this. Does Toolchain already offer that break down by lint tool workunit?

here is a smaller example of a

StreamingWorkunits

plugin: the interface is

WorkunitsCallback

: https://github.com/pantsbuild/pants/blob/dfbfd75f3b1ce1d4c53a639eaeda2254d75078d8/[…]stprojects/pants-plugins/src/python/workunit_logger/register.py

okay cool good to know we can get this data. Thank you for your support on this it's very helpful

For a one-off investigation, a short-cut that avoids a plugin might be something like:

./pants --stats-log --no-local-cache --no-pantsd lint --only=isort ::

(and similarly with

--only=flake8

), and then looking at the

local_process_total_time_run_ms: ...

. I think this will include the time to set-up the tools, but if you're just doing this to work out why linting is taking a very long time (or similar), I suspect that it'll be enough to narrow down the problem