make it compatible with new hashing

[narimiran]

This is now the corrected version: the new hashing algo was slightly tweaked and it won't be part of Nim v1.0.2, so the modifications are needed only for devel (v1.0.99) and for future minor versions (v1.1, etc.).

[def-]

lgtm, you could also sort the results and then compare so that the comparison stays the same.

I'm not sure I understand why the order changes btw, could you explain?

[narimiran]

I'm not sure I understand why the order changes btw, could you explain?

Before hash(123) == 123 (for any int its hash was that int), plain and simple. The problem was when collisions start to occur in HashSet or HashTable, because every hash is bitmasked with the container length, so different (consecutive) integers might want to go into the same bucket, over and over and over again, which caused huge slowdowns.

The fix proposed in nim-lang/Nim#12407 makes hash(123) == 123 * 11, which hugely helps with avoiding collisions, making everything three orders of a magnitude faster! The consequence of this different hashing algorithm is that keys are now not sorted the same as they were before.

you could also sort the results

This is probably a better solution, as one should not depend on some internal order of things.

[def-]

Aaah, finally I understand, it's because of the "let d = {1: 4, 2: 5, 3: 6}.toTable". Better fix would be to make that have deterministic order. Do you want to do so or should I? I think values(d) could just be [4, 5, 6] instead.

[narimiran]

Do you want to do so or should I?

I think we will save at least one iteration if you do it :)

[def-]

Done.