Wrapping and unwrapping DataNewtype blows up script budget

[colll78]

plutarch-plutus/plutarch-ledger-api/src/Plutarch/LedgerApi/V1/Time.hs

Lines 67 to 96 in 429d1b0

	instance PIntegral PPosixTime where
	{-# INLINEABLE pdiv #-}
	pdiv = phoistAcyclic $ plam $ \t1 t2 ->
	pposixTime (pdiv # unPPosixTime t1 # unPPosixTime t2)
	{-# INLINEABLE pmod #-}
	pmod = phoistAcyclic $ plam $ \t1 t2 ->
	pposixTime (pmod # unPPosixTime t1 # unPPosixTime t2)
	{-# INLINEABLE pquot #-}
	pquot = phoistAcyclic $ plam $ \t1 t2 ->
	pposixTime (pquot # unPPosixTime t1 # unPPosixTime t2)
	{-# INLINEABLE prem #-}
	prem = phoistAcyclic $ plam $ \t1 t2 ->
	pposixTime (prem # unPPosixTime t1 # unPPosixTime t2)
	-- | @since 2.0.0
	instance PNum PPosixTime where
	{-# INLINEABLE (#*) #-}
	t1 #* t2 = pposixTime (unPPosixTime t1 #* unPPosixTime t2)
	{-# INLINEABLE (#+) #-}
	t1 #+ t2 = pposixTime (unPPosixTime t1 #+ unPPosixTime t2)
	{-# INLINEABLE (#-) #-}
	t1 #- t2 = pposixTime (unPPosixTime t1 #- unPPosixTime t2)
	{-# INLINEABLE pabs #-}
	pabs = phoistAcyclic $ plam $ \t -> pposixTime (pabs # unPPosixTime t)
	{-# INLINEABLE pfromInteger #-}
	pfromInteger = pposixTime . pconstant
	{-# INLINEABLE pnegate #-}
	pnegate = phoistAcyclic $ plam $ \t -> pposixTime (pnegate # unPPosixTime t)
	{-# INLINEABLE psignum #-}
	psignum = phoistAcyclic $ plam $ \t -> pposixTime (psignum # unPPosixTime t)

This module is bricked. Switched from PInteger to PPosixTime and my benchmarks had a huge regression. The issue is that all the operations wrap and unwrap the data, this results in a ton of iData and unIData builtin calls, because each addition unwrapped each integer from data and then wraps the result. Instead we should have two definitions PosixTimeData and PosixTime or anything else so that this can be avoided and we can just keep it in the non-data encoded format until we want to go back to data encoding.

[SeungheonOh]

Yes, I'm aware of this issue. We are going to remove PDataNewtype which was a big mistake. It made types more intuitive, but really obfuscated cost of construting and deconstructing data value