Anyone have thoughts on allowing

./pants run <tool_subsystem>

? I want to know what args

black

or

pylint

accepts. Running

./pants run black -- --help

would be AMAZING (Also this could possibly replace

repl

, if we support the REPLs in

run

. Huzzah less goals? I could do this with my own separate

pex_binary

and resolve, but I think we all understand how nasty that it 🙂

I could do this with my own separate

pex_binary

and resolve, but I think we all understand how nasty that it 🙂

fairly nasty, yea. but depends on the usecase… if your goal to be identical to what pants does under the hood when sandboxed (identical config detection, identical options, etc) then there is other syncing of options that you’d need to do. if your goal isn’t to be identical to what we do in the sandbox, then you don’t necessarily need an identical version

Having to duplicate the reqs in my env and re-lock is prohibitively annoying

also only works if the linter doesn’t need thirdparty deps

Yeah... actually

./pants lint --only=black --black-args='--help'

might work (no specs)? But is cumbersome Also I'm still thinking about

./pants run [i]python

as a replacement for

repl

for Python

i’m still not sure that i understand the

black

usecase.

It's just an example of a tool I want to run. A better example is

pytest

pytest definitely wouldn’t have thirdparty deps though.

E.g. what's the pytest arg to filter to a subset of tests

you’d need awareness of which targets are being run to put the appropriate deps in a sandbox

Of course, in this case

--help

s a useful use-case

Before diving in, I just want to observe this (ok, rather awkard) cludge:

$ ./pants -ldebug fmt --only=“black” --black-args=--help src/python/pants/bin:bin

Yeah I point that out above, but also: • eww • Now do

pytest

😉 • Now do

ipython

😉

Ah, right (didn’t read the whole thread 😛 )

I'm kinda also hitting this from the (upcoming) Jupyter notebook support. I'ts kind of like a runner and kind of like a repl. I'm kinda wondering if we could run our tools through

run

, would

repl

be as necessary? (We'd need to figure out third party dep handling, but lets not let implementation details stop progress 🙂 )

This one actually touches slightly on this, I think: https://github.com/pantsbuild/pants/pull/13257

Yeah I think they are in the same vein

Another thing to consider is that dependencies currently only are targets, but what if we could have a dependency on the digest of a tool, then we could populate it into a sandbox for a experimental_shell_command target and do whatever.