#general
- m
mammoth-pharmacist-54215
08/20/2015, 7:06 PMany lang that uses a repl has an interpreter though, and repl has to know which interpreter to load based on the target. So I’m not sure why this can’t be used to override the default on a global scope… it doesn’t logically make sense to run repl when multiple types would be used… something like ./pants repl :: is nonsensical if there are scala and python projects. - b
bored-art-40741
08/20/2015, 7:10 PMIt's not really a matter of what makes sense in the context of that task, it's just how the options system works - b
bored-art-40741
08/20/2015, 7:10 PMIt is not capable of doing that kind of disambiguation, by design - m
mammoth-pharmacist-54215
08/20/2015, 7:14 PMCurrently the system picks the interpreter based on lang and overrides, so why not have the option have something like <DEFAULT> as a default. If that is set the system would use current behavior, and then if it extists wouldn’t the individual repl implementation that normally triggers either python or scala be able to look in the appropriate context for that value and if its not <DEFAULT> then use the intepreter specified? - b
bored-art-40741
08/20/2015, 7:16 PMOptions are only scoped to goals when the task that registers the option has the same name as the goal. A lot of thought went into that decision based on getting burned in the past where "helpful" disambiguation led to very difficult to diagnose bugs when tasks were added to or removed from a goal. I do not anticipate that behavior changing, and a
task doesn't really make any sense.repl.repl - b
bored-art-40741
08/20/2015, 7:17 PMThis is a super minor point, but the point still stands: that's how the scoping will work on the option. - b
bored-art-40741
08/20/2015, 7:17 PMShort of introducing a subsystem for this concept, which might make sense, but I haven't put much thought into it - f
fancy-queen-20734
08/20/2015, 7:18 PMThe thing I’ve been wanting for options is to have option look up failures cause a option parser tree walk down scope that provides a good error message suggesting a set of scoped options. - f
fancy-queen-20734
08/20/2015, 7:18 PMso
would say “did you mean —repl-python-interpreter?./pants repl —interpreter=ipython … - b
bored-art-40741
08/20/2015, 7:19 PMProviding better error messages is always laudable, that sounds feasible - m
mammoth-pharmacist-54215
08/20/2015, 7:19 PM@bored-art-40741: fair enough, subsystems may be able to resole this for me… I’ll take a look. So far my take away is that people don’t object to adding an option, but switching to ipython by default people do object to. - b
bored-art-40741
08/20/2015, 7:19 PMYeah. Benjy also wants to clarify the terminology I think, and I think he's right that we should be sure about what we mean by "interpreter" - f
fancy-queen-20734
08/20/2015, 7:19 PMit could make options migrations less painful by pointing to the new correct option - h
happy-kitchen-89482
08/20/2015, 7:46 PMCurrently "interpreter" means something slightly different than "shell". - h
happy-kitchen-89482
08/20/2015, 7:46 PMYou can control the interpreter selection today via the PythonSetup subsystem - h
happy-kitchen-89482
08/20/2015, 7:46 PMthat is true for all PythonTasks - h
happy-kitchen-89482
08/20/2015, 7:46 PMSo whatever we do here, let's not call it --interpreter - h
happy-kitchen-89482
08/20/2015, 7:46 PM--shell is fine - m
mammoth-pharmacist-54215
08/20/2015, 8:01 PMokay, great thanks - m
mammoth-pharmacist-54215
08/21/2015, 12:42 AMif anyone has a spare moment would they mind looking at: https://rbcommons.com/s/twitter/r/2593/ - e
enough-analyst-54434
08/21/2015, 3:12 AMmutable conversation streams are truly odd! - h
happy-kitchen-89482
08/21/2015, 3:12 AMIndeed! - h
happy-kitchen-89482
08/21/2015, 3:13 AMAll I wanted to do was show task subsystem help alongside the task help - h
happy-kitchen-89482
08/21/2015, 3:13 AMit has turned into an epic yak shave - e
enough-analyst-54434
08/21/2015, 3:16 AMAre the Task/Subsystem dependencies getting in the way? When I spiked on that change a few weeks back there was some knot that only presented itself in ~10 tests as I recall, and the knot was too hard to grok or too hairy at the time for a quick spike - h
happy-kitchen-89482
08/21/2015, 3:17 AMI don't think so - the issue is the interplay between recursive options, inherited option values and re-registering options on inner scopes. - h
happy-kitchen-89482
08/21/2015, 3:18 AMBasically: Currently we register subsystem options only on its global scope (e.g.,
) and for options to be changed on task-specific scopes (e.g.,jvm
) they must be recursive=True.jvm.scala.repl - e
enough-analyst-54434
08/21/2015, 3:18 AMIs the DEFAULT 4th axis of un-intended inheritance getting in your way? - h
happy-kitchen-89482
08/21/2015, 3:18 AMNo, DEFAULT seems to be fine here. - e
enough-analyst-54434
08/21/2015, 3:18 AMk