Ergonomic, hermetic, user-friendly software build system for Python, Go, Java, Kotlin, Protobuf, Scala, Shell. Pants lets you fearlessly scale up your codebase!

Pants

Yeah, I’m about to do that, but it seems like a hacky way to do things.

<@U051221NF>, <@U053HNE90>: would you guys object to a patch that moved JvmDependencyAnalyzer out to its own task, which consumes analysis as a product?

is there a shorter unique id to use for targets than `target.id`? a hash maybe?

we're running into classpath length issues which will likely be solved by <https://github.com/pantsbuild/pants/issues/1988> in the long term, but which in the medium term would be avoided by using shorter ids

Maybe like, target subsystem with an option to replace target.id with a short-hash of the target id?

I think we've been slowly creeping toward a target subsystem as it is

generally? hm, yea... that would be neat. with a cache key gen bump prolly.

As far as Jvm dep analyzer, that's primarily a benjy question.  It sounds reasonable to me, and as Benjy says: we've never regretted moving something into a product

does anyone recall the reasoning behind the prior crusade against code in __init__.py files? just unnatural to crack open a __init__.py as the main code location for a given module?

<@U054MQUPL> plugins / contribs and fear of the dreaded mis-matched __init__.py contents for the same package, also I compounded issues with earyly aping of a python git library that did alot of aliasing in __init__.py that could lead to funky issues depending on what imported 1st.

So, even though you can do code right, it was deemed to dangerous for too little benefit

so we live with pants.java.distribution.Distribution type names

how does python_eval avoid sideeffects on import?

you can't without advanced sandboxing technology

if you write code that has side effects we just helped you find out (DB nuked!) ?

This is part of the reason its not installed by default

I imagine any lang with an eval is hard to sandbox securely with just ASTs

I'm not sure how goog did it for appengine