Dynarray.{equal, compare}

[gasche]

This was suggested by @OlivierNicole in #12885 (comment) .

Having comparison functions for Dynarray is useful because they do not have a canonical representation: two dynarrays may have the same elements, but yet have (=) return false or Stdlib.compare return non-0 due to having a different capacity.

I reused and adapted the docstring of List.equal and Seq.compare.

Comparing of different-length arrays

For compare there is a choice between:

1. what wikipedia calls the lexicographic order: [10] > [1; 2] because 10 > 1
2. what wikipedia calls the "shortlex" order: [10] < [1; 2] because the first list is shorter (before comparing the elements)

Using Stdlib.compare (Dynarray.to_list a) (Dynarray.to_list b) would result in lexicographic order (1), and Stdlib.compare (Dynarray.to_array a) (Dynarray.to_array b) would result in shortlex order (2). (In other words, "doing the same as other data structures" does not force us to choose one or the other.)

I went for the shortlex order which is slightly faster.

[dbuenzli]

I just left bikesheedding comments on the docstrings.

[gasche]

Thanks both, I took your comments into account.

[yallop]

Having comparison functions for Dynarray is useful because they do not have a canonical representation: two dynarrays may have the same elements, but yet have (=) return false or Stdlib.compare return non-0 due to having a different capacity.

Extending this reasoning, we should have comparison functions for other "container" types in the standard library, too, because their elements might not have canonical representations. For example, Array.equal is also useful, because (=) might return false for an int Dynarray.t Array.t comparison where corresponding Dynarray.t elements have the same int elements but different capacities.

[gasche]

I don´t disagree and in fact I was surprised to find that Array does not have those functions. (We have equal functions for base types thanks to #84.) But I would prefer to restrict the scope of the present PR to Dynarray comparisons only to avoid getting stuck.

[OlivierNicole]

This looks good to me.

[OlivierNicole]

Smart!

[gasche]

(... and only correct because we know that they are both non-negative, I think.)

[OlivierNicole]

Well, insofar as you consider that a dynarray of length −12 is lesser than a dynarray of length −2, it is correct!

[shindere]

Olivier Nicole (2024/05/16 08:45 -0700):

Well, insofar as you consider that a dynarray of length −12 is lesser than a dynarray of length −2, it is correct!

:-)