Using large records with `Barbies` leads to extreme comptime memory use

Learn
1

Suppose I have a large record (with 20+ fields), and I’d like to use HKD on it. But when I compile the following code, it uses up all my 8G memory before being terminated. (Compiled with GHC-9.8.2, on WSL2-Ubuntu-22.04)

module Test where

import Prelude
import GHC.Generics
import Barbies
import Barbies.Bare
import Data.Kind

data Foo t (f :: Type -> Type) = Foo
  { field01 :: Wear t f Int
  , field02 :: Wear t f Int
  , field03 :: Wear t f Int
  , field04 :: Wear t f Int
  , field05 :: Wear t f Int
  , field06 :: Wear t f Int
  , field07 :: Wear t f Int
  , field08 :: Wear t f Int
  , field09 :: Wear t f Int
  , field10 :: Wear t f Int
  , field11 :: Wear t f Int
  , field12 :: Wear t f Int
  , field13 :: Wear t f Int
  , field14 :: Wear t f Int
  , field15 :: Wear t f Int
  , field16 :: Wear t f Int
  , field17 :: Wear t f Int
  , field18 :: Wear t f Int
  , field19 :: Wear t f Int
  , field20 :: Wear t f Int
  }
  deriving Generic

instance FunctorB (Foo Bare)
instance FunctorB (Foo Covered)
instance ApplicativeB (Foo Covered)
instance ConstraintsB (Foo Bare)
instance ConstraintsB (Foo Covered)
instance BareB Foo

I wonder if it is an intended behavior or a memory leak. If the former is the case, are there any alternative methods to do similar things?

Any suggestions are appreciated.

2 Likes
2

Oh, I found this issue may be related to the problem.

And currently the simplest way is to set optimize level to -O0.

3

I haven’t used Barbies but I filed a issue a while ago when writing code which I believe is similar to barbies use case (trying to sequence a record MyRecord f to f (MyRecord Identity) which you can write manually using lots of <*> (as in `MyRecord <$> field1 <> field2 <> … <*> fieldn).

I though the bug was eventually fixed (but I’m not sure as I found a work around). It might be related (or not).