BTreeMap: lightly refactor range_search

[ssomers]

Ramp up to #81094 but IMHO useful in itself.

r? @Mark-Simulacrum

[Mark-Simulacrum]

Can we place a comment on what the meaning of Some/None here is?

[ssomers]

How about a commented type?

@rustbot modify labels: -S-waiting-on-author +S-waiting-on-review

[Mark-Simulacrum]

Looking a bit more at this, I am struggling with seeing this as an improvement. It seems like the encoding of the bit true/false into the Option makes it less clear what the code is doing rather than more by combining the paths for initial unbounded with "became unbounded" in a way that's confusing to me at least.

Maybe we can find a different structure that works well? Maybe noting on the various branches why first/last/left/right edges are chosen would also help with making the intent clear, right now I at least have a bit of a hard time justifying without some manual working out why they're being made.

[ssomers]

I can only repeat I think it's slightly better in itself and that #81094 is the next step, which splits this large function up, and is also necessary for the drain proposal. If you have something concrete to alter, I can do that. If you want to skip this step, fine by me. If you want to skip the rest of drain, that's also fine.

[ssomers]

The only thing I can think of that doesn/t thwart #81094 is the definition of NoneableBound: name, whether it is a proper enum, reusing Bound or redefiningm perhaps more clearly distinguishing "include none" from unbounded.

[ssomers]

Can't keep updating this preparation of a preparation