Rewrite rules

[L-as]

We want rewrite rules to support things like #128.

Rewrite rules should function like GHC's rewrite rules, albeit they should fire consistently. This is easier for us
to do because of the lack of complex optimisations.

[delicious-lemon]

Yes! Is there any work toward this ongoing?

My motivation is https://github.com/mlabs-haskell/apropos-tx/blob/26294dbaa3a1c9e794bdeab2af98d2f74624ab96/src/Apropos/Script/Iso/Arrow.hs#L49

These combinators introduce unnecessary construction/deconstruction of builtin pairs which could be reduced easily by a rewrite.

[L-as]

I am working on this internally.

[srid]

• Having a rewrite rule to "drop" traces (as part of Replace development flag with runtime compile options #359) which incidentally resolves Remove flakiness in fail pattern match messages #290

[blamario]

Just to be clear, are the rewrite rules supposed to operate on the compiled Scripts like Shrinker does, or on a Term?

[L-as]

@blamario They operate on the precompilation AST.

[L-as]

I think I have a viable design FWIW

[srid]

What sort of optimizations will rewrite rules provide?

For instance, if I use pfield @"txInfo" ctx in more than one place, will that get optimized (ie, hoisted at top-level and gets used via reference, instead of recomputing)?

[L-as]

That's general CSE and not something I have tried to implement in Plutarch.
It's not as easy as it might seem, because terms can err.
if you have two equivalent terms t and t', and t' is only evaluated if t is, then you can in fact
use t for both, but we'd have to detect that somehow.

[L-as]

In some places however this means we have to do a plet and that's not free.

[L-as]

They're supposed to operate on Plutarch terms.