Add deriveEq for Plinth similar to deriving stock Eq

[bezirg]

Context

• What: Add Template Haskell deriveEq function to automatically generate PlutusTx.Eq instances
• Why: Eliminate boilerplate and reduce errors from manual Eq instance implementation across plutus-ledger-api
• Issue: Closes Add TH derivings for Plinth's Eq #7430

Approach

Implemented a Template Haskell function that generates structural equality similar to GHC's deriving stock Eq. The implementation:

• Refactored module structure: Split PlutusTx.Eq into PlutusTx.Eq.Class (type class definition) and PlutusTx.Eq.TH (deriving function) for better organization
• Smart variable naming: Generates unique variable names (e.g., l1l, r2r) to avoid shadowing warnings
• Phantom type support: Uses role checking to exclude phantom type parameters from instance constraints, allowing derivation for types like PhantomADT e
• Short-circuit evaluation: Generates efficient code that stops at the first inequality
• INLINEABLE pragmas: Ensures cross-module optimization for on-chain code

The generated code handles:

• Multi-constructor ADTs with catch-all clause for different constructors
• Single-constructor types without unnecessary catch-all
• Empty types (Void) with trivial True equality
• Records and regular constructors uniformly

Changes

Template Haskell Implementation

• plutus-tx/src/PlutusTx/Eq/TH.hs (NEW): Core deriveEq function using th-abstraction for reliable type introspection
• plutus-tx/src/PlutusTx/Eq/Class.hs (NEW): Extracted Eq class definition from main module
• plutus-tx/src/PlutusTx/Eq.hs: Refactored to re-export from submodules

Applied Across plutus-ledger-api

Replaced ~100 lines of manual Eq implementations with deriveEq calls in:

• V1: Address, Credential, DCert, Tx, Value, Interval, Contexts
• V2: Contexts, Tx
• V3: Contexts, MintValue, Tx

Net result: -101 lines (602 deletions, 501 additions) primarily from removing boilerplate

Tests and Golden Files

• plutus-tx/test/Eq/Spec.hs (NEW): Comprehensive test suite covering:
  • Multi-constructor ADTs with mixed field types
  • Newtypes
  • Records
  • Phantom types
  • Void types
  • Short-circuit evaluation behavior
• Golden tests: TH code generation snapshots for SomeLargeADT and PhantomADT

Updated Benchmarks

• Regenerated golden PIR/UPLC files reflecting new variable naming scheme

[github-actions]

Execution Budget Golden Diff

670e913 (master) vs 3b9aaed

output

plutus-benchmark/cardano-loans/test/9.6/main.golden.eval

Metric	Old	New	Δ%
CPU	111_100_889	110_972_889	-0.12%
Memory	622_350	621_550	-0.13%
Flat Size	8_663	8_660	-0.03%

plutus-benchmark/coop/test/9.6/mustBurnOwnSingleton.golden.eval

Metric	Old	New	Δ%
CPU	114_113_035	114_049_035	-0.06%
Memory	579_805	579_405	-0.07%
Flat Size	3_833	3_830	-0.08%

plutus-benchmark/linear-vesting/test/9.6/main.golden.eval

Metric	Old	New	Δ%
CPU	30_837_131	30_405_131	-1.40%
Memory	131_619	128_919	-2.05%
Flat Size	2_860	2_351	-17.80%

plutus-benchmark/script-contexts/test/V3/Data/9.6/purposeIsWellFormed-4.golden.eval

Metric	Old	New	Δ%
Flat Size	1_839	724	-60.63%

This comment will get updated when changes are made.

[SeungheonOh]

I'm not too confident in my ability to review TH code. So perhaps a look from @Unisay would be nice. But everything looks right from what I can tell.

[SeungheonOh]

Also, can you check if this works with polymorphic phantom types? I'm curious as per #4537

[Unisay]

Perhaps in another PR?

If a PR is useful improvement even without its follow-up I'd say yes.

[ana-pantilie]

Approving because my comments aren't critical, but please take a look at them if you can.