Update Bitmap::len to return bits rather than bytes

[matthewmturner]

Which issue does this PR close?

Closes #730

Rationale for this change

What changes are included in this PR?

Are there any user-facing changes?

[matthewmturner]

@jorgecarleitao FYI. I did what i thought was the simplest update by simply multiplying bytes by 8 and updating Bitmap tests.

when running all arrow-rs tests locally i had a number of errors on ipc reader and writer tests - not sure if its related. i will look into it more though.

let me know your thoughts :)

[jorgecarleitao]

Looks a good start. I think that every bitmap.len() out there needs to be divided by 8 now, since this is a backward-incompatible change.

[matthewmturner]

The errors were actually just because i needed to add the submodules.

I see you already approved - do you still need me to check if bitmap.len is used anywhere?

[jorgecarleitao]

No, checks pass, it should be good to go.

Thank you for your contribution!

[alamb]

Thank you @matthewmturner -- I think this does what is called for in #744 👍

@jorgecarleitao perhaps you can have a look as well?
(edit -- missed that Jorge had already reviewed)

I also looked at the uses of Bitmap::len in the arrow crate and it only appears to be used in tests:

/Users/alamb/Software/arrow-rs/arrow/src/bitmap.rs
120:         assert_eq!(64, Bitmap::new(63 * 8).len());
121:         assert_eq!(64, Bitmap::new(64 * 8).len());
122:         assert_eq!(128, Bitmap::new(65 * 8).len());

One concern I have is related to backwards compatibility -- if we change len to return the length in bits but change nothing else, the rustc compiler will not alert any users of this function that the API has changed.

What about also renaming Bitmap::len() to Bitmap::byte_len() to make it super clear what is going on as well as generating compiler errors on upgrade?

[alamb]

Can we also add a docstring here that explains what Bitmap::len actually returns (aka the length in bits, not bytes)?

[jorgecarleitao]

I would not consider this backward incompatibility but a bug fix: imo Bitmap::len should be in bits, not bytes. We actually found this while assuming that the len was in bits (it is a bitmap, after all).