Let ApInt::resize_to_bool just return the least-significant bit (should be faster)

[Robbepop]

[AaronKutch]

There are the following least and most significant functions right now:

• most_significant_bit (private)
• least_significant_bit (private)
• sign_bit, and set and unset and flip
• is_even
• is_odd
• is_negative
• is_positive

and related to these are:

• from_bool and the From<bool>
• resize_to_bool
• try_to_bool

I will check that these are optimized, but should I remove some of these (e.g. is_negative should only be for Int or should it also be for ApInt?)

[AaronKutch]

I was looking at the is_one function and remembered that ApInts with bitwidth 1 treated as signed have only a sign bit, meaning that a bit value of 1 is treated as -1. There was a bug caused by a similar thing I dug up here: https://github.com/rust-lang/rust/issues/51582. Should I make a is_unsigned_one and is_signed_one and add some documentation about the corner case?

[AaronKutch]

Maybe I should let the current functionality be, specify that "Returns true if the unsigned interpretation ApInt represents the value one (1).", and put more general documentation higher up about bitwidth 1 ApInts.

[AaronKutch]

Reminder to myself that advanced slice matching is going to be stabilized, so I can use stuff like

fn ls_digit<T>(x: &[T]) -> Option<&T> {
    match x {
        [x, ..] => Some(x),
        [] => None,
}

[AaronKutch]

Even though I usually like verbose member functions, I think we want to use hungarian notation for these cases of accessing the most and least significant bits: ApInts will have public member functions:

• lsb for least significant bit
• msb for most significant bit
• set_lsb for setting the significant bit
• set_msb all the others will be removed.

Maybe Ints will have in addition to the above:

• sign_bit
• is_negative
• is_positive

[Robbepop]

Sounds good about lsb and msb. Not sure about getting rid of is_even and is_odd since they carry more meaning than a check to lsb etc. Depends on what we want Apint to be. Do we want it to be a user-friendly type or do we want it to be a generalist type that should be embedded into wrappers. For example the stevia crate for which I was implementing this crate mainly used it in the latter case. So not directly, but wrapped it providing its own interfaces on top.

[AaronKutch]

I think there should be one standard way of getting the least significant bit. When dealing with regular integers, I always think of it in terms of the least significant bit.

[AaronKutch]

Actually, we can keep is_even and is_odd