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

Pants

<@U066LC126> : so, it looks like the typeids on Values got hella expensive with pickling for ids

i'm going to start that now. i think it shouldn't shake things up too much outside of `<http://externs.rs|externs.rs>`/`native.py`

<https://github.com/pantsbuild/pants/pull/4274>

<@U066LC126> : about 30% of runtime is spent in `store_list` while merging HydratedTargets up to the root... i'm getting very tempted to make the result of `SelectDependencies` a rust Vec

so that we can merge them all at once instead

on the fence about tackling that or landing the io first

also very likely to be useful to move storage of `Key`s to rust... basically, you'd pickle and fingerprint on the rust side, but then store a mapping from `Key`-&gt;`Value` on the rust side... that looks like it could be about 10-15%

the cool thing about most of these optimizations is that they don't make life more difficult... easier, if anything

<@U066LC126> : do you have plans to overhaul RuleGraph currently? or is the next step just the node generation

i'm going to schedule a meeting for tomorrow afternoon cause i feel like we're in the homestretch on this, but need to sync.

<@U066LC126> , <@U098ZNTCN> : re-ping on <https://github.com/pantsbuild/pants/pull/4274>

it’s dropping the handle to the first item in the tuple?

<@U054MQUPL> : `pub struct Value(Handle)` is a "tuple struct"

<@U066LC126> : one concrete question for you (can respond on the PR if you want): how should i handle the `Snapshot` and `FilesContent` hacks i have in place here: <https://github.com/pantsbuild/pants/pull/4265/files#diff-11db197014e1e80c9551e9b61aba6aa0R409>

currently it's implemented by installing empty/noop tasks on the python side, and then replacing all requests there.