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

Pants

(I guess a certain mix of rglobs and globs would allow that by making code available in multiple `python_library`)

<@U43EQKGHW> : nesting BUILD files is perfectly fine. but you have to explicitly depend on the targets defined in those BUILD files if you expect to be able to import them

<@U06A03HV1> scalatest or specs2? :slightly_smiling_face:

having multiple targets owning the same code is generally not a good idea, as everything happens at the target level, so everything will be executed/cached multiple times

<@U07QGT3V3> , <@U34SFN1K8> : note that scalafmt also supports autoformatting! we haven't taken the plunge internally, but you should be able to (*carefully!*) run `./pants fmt ::` to have it fix up code

:thumbsup: we have our own style guide and don't have time to write a corresponding format ATM but we're thinking about it for the future

we have six+ internal styles, and are hoping to standardize on "scalafmt defaults". we'll see.

does 18 seconds to compile 6 scala sources seem right? :thinking_face:

Back in the day at 4sq we were averaging 1 second per .scala source file, but obviously it depends how big they are and how much implicit resolution and type inference they require.

Also, we found that the compiler can be IO bound

It emits so many classfiles per source file...

<@U051221NF> what’d u do to get down to that perf? we’re far worse then that. we’re pretty good on implcits for example we only allow private implicits, so they are always contained in a single class/file

we have no implicits in our own code, only external libs like json4s

<@U0DJ6JX08> we gave every engineer a 16GB stick of RAM and a screwdriver.

And I guess our files were on average smaller than yours?

The worst offenders were codegen’d scala, but the artifact cache saved us there

Custom generator, but off .thrift files, yes