Adding `.reverse()` method to `IndexMap` & `IndexSet`

[linclelinkpart5]

Looking at the impl of OrderMapCore, .sort_by() saves the hash index, sorts the underlying entries vector, and then restores the hash index. Would this mean that a .reverse() method could be added to OrderedMapCore (and thus IndexMap/IndexSet)?

[cuviper]

Reverse would be possible, yes. In fact, we can probably do the fixup more efficiently than sort since the change is predictable -- just sweep the hashed indices and flip each to len - index.

[linclelinkpart5]

Reverse would be possible, yes. In fact, we can probably do the fixup more efficiently than sort since the change is predictable -- just sweep the hashed indices and flip each to len - index.

Just to understand correctly, this manipulation of the indices would replace the saving/restoring of the hash index that .sort_by() does normally, correct?

I created the implementation, but tests are failing right now, and I'm not exactly sure why!

[cuviper]

Just to understand correctly, this manipulation of the indices would replace the saving/restoring of the hash index that .sort_by() does normally, correct?

Yes.

I created the implementation, but tests are failing right now, and I'm not exactly sure why!

Can you share the code and details on what's failing?

[linclelinkpart5]

Hey @cuviper, thanks for the info! I realized the cause of my test failures: I forgot that when inserting multiple key-value pairs into an IndexSet, any duplicated keys stay put, and only their values get updated. I was wrongly assuming that adding a duplicated key brought that KV pair to the right, order-wise.

I've created a PR, but it would still need the improved re-indexing you suggested.