haskellari/strict-containers

strict-containers's Issues

Would it be possible to port to GHC 9.10?

The changes would probably be similar to what's in

haskell/vector#462

See

haskell/core-libraries-committee#167

for context and the breakage got revealed in

https://gitlab.haskell.org/ghc/ghc/-/issues/23848#note_527774

Thank you!

GHC 9 regression

Filed as https://gitlab.haskell.org/ghc/ghc/-/issues/19700

review HashSet

containers' Set is defined as a separate data type where every field is strict so it is obvious there is no value to defining a "more strict" one here, since it would be exactly the same.

OTOH, unordered-containers' HashSet is defined as a newtype wrapper around HashMap _ (). Because HashMap internally uses a primitive Array# which is lazy in its contents, this means that HashSet may contain thunks even when the values are all (). This could be what's behind haskell-unordered-containers/unordered-containers#70.

If we instead define HashSet from our stricter HashMap, whose array operations all force the value, then we can avoid this potential problem.

Add to stackage

Thanks for this package. It would be convenient for to have it on Stackage.

Could you open an MR with Stackage to add it to the package set?

I understand that this adds a further maintenance burden, so if you'd like I can maintain the Stackage entry and open MR's here when bounds need to be updated, etc.

Would it be possible to port to GHC 9.6?

I wonder, perhaps relaxing the strict bound would suffice?

Seq: make whole/part-container transforms such as fmap strict

As described in the latest documentation. Possible candidates beyond fmap:

• scan*
• update, insertAt, and anything else we might've missed with a similar type signature -> a .. -> (Seq a, *).
• dynamic constructors e.g. from*, replicate*, iterateN. although the documentation says they are O(n), from a quick reading of the source code it seems this really means "only if someone actually evaluates the whole thing e.g. by printing it", since the lazy parts of the Deep structure stay lazy

Some technical details in haskell/containers#752

Doesn't compile with GHC 9.4.2 due to the upper bound on hashable

Adding --allow-newer=strict-containers:hashable makes it compile. Without it, I'm getting

[__1] trying: strict-containers-0.1 (dependency of horde-ad)
[__2] next goal: hashable (dependency of strict-containers)
[__2] rejecting: hashable-1.4.1.0 (conflict: strict-containers =>
hashable>=1.2.7.0 && <1.4)
[__2] skipping: hashable-1.4.0.2, hashable-1.4.0.1, hashable-1.4.0.0 (has the
same characteristics that caused the previous version to fail: excluded by
constraint '>=1.2.7.0 && <1.4' from 'strict-containers')
[__2] trying: hashable-1.3.5.0
[__3] next goal: base (dependency of horde-ad)
[__3] rejecting: base-4.17.0.0/installed-4.17.0.0 (conflict: hashable =>
base>=4.5 && <4.17)

Document performance bounds & laziness of Data.Strict.Sequence

The performance of Data.Sequence is tied together with its laziness, so the strict version might confuse some people on what the performance is and how lazy it is. We should document it.

Doesn't compile with vector-0.13 (with --allow-newer in order to ignore the bound from .cabal)

Compiles fine with vector-0.12.3.1. The failure is

[41 of 44] Compiling Data.Strict.Vector.Autogen.Mutable ( src/Data/Strict/Vector/Autogen/Mutable.hs, dist/build/Data/Strict/Vector/Autogen/Mutable.o, dist/build/Data/Strict/Vector/Autogen/Mutable.dyn_o )

src/Data/Strict/Vector/Autogen/Mutable.hs:161:3: error:
    Data constructor not in scope:
      INTERNAL_CHECK :: t2 -> String -> String -> Bool -> m2 () -> m ()
    |
161 |   INTERNAL_CHECK(check) "moveBackwards" "not a backwards move" (dstOff < srcOff)
    |   ^^^^^^^^^^^^^^

src/Data/Strict/Vector/Autogen/Mutable.hs:161:18: error:
    Variable not in scope: check
    |
161 |   INTERNAL_CHECK(check) "moveBackwards" "not a backwards move" (dstOff < srcOff)
    |                  ^^^^^

src/Data/Strict/Vector/Autogen/Mutable.hs:168:3: error:
    Data constructor not in scope:
      INTERNAL_CHECK :: t1 -> String -> String -> Bool -> m1 () -> m ()
    |
168 |   INTERNAL_CHECK(check) "moveForwardsSmallOverlap" "not a forward move" (dstOff > srcOff)
    |   ^^^^^^^^^^^^^^

src/Data/Strict/Vector/Autogen/Mutable.hs:168:18: error:
    Variable not in scope: check
    |
168 |   INTERNAL_CHECK(check) "moveForwardsSmallOverlap" "not a forward move" (dstOff > srcOff)
    |                  ^^^^^

src/Data/Strict/Vector/Autogen/Mutable.hs:179:3: error:
    Data constructor not in scope:
      INTERNAL_CHECK :: t0 -> String -> String -> Bool -> m0 () -> m ()
    |
179 |   INTERNAL_CHECK(check) "moveForwardsLargeOverlap" "not a forward move" (dstOff > srcOff)
    |   ^^^^^^^^^^^^^^

src/Data/Strict/Vector/Autogen/Mutable.hs:179:18: error:
    Variable not in scope: check
    |
179 |   INTERNAL_CHECK(check) "moveForwardsLargeOverlap" "not a forward move" (dstOff > srcOff)
    |                  ^^^^^
Error: cabal: Failed to build strict-containers-0.1 (which is required by
test:shortTestForCI from horde-ad-0.1.0.0, test:minimalTest from
horde-ad-0.1.0.0 and others). See the build log above for details.

add tests

Would be good to copy the tests over here too and run them