Array.append causes runtime error on largish arrays

[klaftertief]

I know there are already some similar Array issues, but I hope to offer a better test case for this specific issue. This is extracted from issues working on a polyline simplification library, for this issue mainly https://github.com/klaftertief/elm-simplify/blob/tail-recursive-array/src/Simplify/RadialDistance.elm.

I applied the open pull requests #396 and #399, but they didn't help.

This test cases crashes in subLengths += a.table[i].table.length (table is undefined) starting at a length of 32^2 - 1*32 + 1.

import Array exposing (append, initialize)
import Graphics.Element exposing (show)

main =
  show <|
    append
      (initialize 1 (always 1))
      (initialize (32^2 - 1*32 + 1) (\i -> i))

It's extracted from this recursive function (which is similar to the one in the linked repo). An interesting thing is that this starts crashing at a length of 32^2 - 2*32 + 1.

import Array exposing (append, empty, get, initialize, length, slice)
import Graphics.Element exposing (show)

run accum array =
  let
    head = get 0 array
    tail = slice 1 (length array) array
  in
    case head of
      Nothing -> accum
      Just x ->
        run
          (append (initialize 1 (always x)) accum)
          tail

main =
  show <|
    run
      empty
      (initialize (32^2 - 2*32 + 1) (\i -> i))

[eeue56]

I spent a while looking at this problem with @klaftertief and sadly was unable to figure out a solid solution. I'm glad you made it an issue! When I get some time I'll look more into this.

[eeue56]

I spent a while looking at this problem with @klaftertief and sadly was unable to figure out a solid solution. I'm glad you made it an issue! When I get some time I'll look more into this.

[klaftertief]

Thanks @eeue56 for your efforts.

This commit fixes the symptoms of this specific issue, see this test.

There are more Array related bugs/crashes popping up in other test cases, but it's hard to find minimal test cases to have a solid foundation to fix those. I hope I'll find some time to add more minimal failing tests and maybe get a better understanding of the native implementation.

[klaftertief]

Thanks @eeue56 for your efforts.

This commit fixes the symptoms of this specific issue, see this test.

There are more Array related bugs/crashes popping up in other test cases, but it's hard to find minimal test cases to have a solid foundation to fix those. I hope I'll find some time to add more minimal failing tests and maybe get a better understanding of the native implementation.

[dan-tull]

I hit this issue today, though for me the Array.append call that is failing is for arrays of length 963 + 1 and 960 + 2 (both slightly lower bounds than the 993 in the example above).

If it would be useful, I could extract minimal test case for the bug in my scenario.

[dan-tull]

I hit this issue today, though for me the Array.append call that is failing is for arrays of length 963 + 1 and 960 + 2 (both slightly lower bounds than the 993 in the example above).

If it would be useful, I could extract minimal test case for the bug in my scenario.

[eeue56]

We're working on it

[eeue56]

We're working on it

[evancz]

Closing in favor of #649 which will track all array issues and progress on a new implementation.

[evancz]

Closing in favor of #649 which will track all array issues and progress on a new implementation.