#11269 Explore replacing monomorphization with implicit conversion Issue created by stuhood We currently "monomorphize" `@rule`s in the

RuleGraph

by creating a distinct copy of a

@rule

per set of

Param

types that it might be used with. This is quite complicated, and a lot of time has been invested in making it "mostly work"... but the implementation is not perfect, and we've needed to apply constraints that can be difficult to reason about. A motivation for monomorphization is that it avoids needing

eq

or

hash

implementations for positional arguments to

@rules

which are computed from `Param`s, rather than being `Param`s themselves. That is of questionable value relative to the complexity it incurs. And additionally, because monomorphization creates additional copies of

@rules

, it's likely that we are memoizing less than we could be. * * * An alternative to monomorphization would be to instead treat the preparation of the positional arguments to a

@rule

as similar to "implicit conversion" (a native concept in Scala, and possible via the

Deref

trait in Rust). This would involve optionally recursing to do "type conversion" via other rules (such as from

Addresses

to

Targets

) before invoking a

@rule

. For the purposes of these conversions, it's likely possible to keep a global lookup table of conversions that are possible, rather than actually encoding the conversion as a dependency of the

RuleGraph

entry. pantsbuild/pants