Re: [PATCH 3/6] rust: add `bitfield!` macro

["Alexandre Courbot" <acourbot@nvidia.com>]

On Wed Jan 28, 2026 at 6:10 AM JST, Gary Guo wrote:
> On Tue Jan 27, 2026 at 9:03 PM GMT, John Hubbard wrote:
>> On 1/27/26 1:57 AM, Alexandre Courbot wrote:
>>> On Tue Jan 27, 2026 at 11:55 AM JST, Yury Norov wrote:
>>> <snip>
>>> 
>>> So while we cannot achieve exactly the short syntax above (which has its
>>> drawbacks as well, such as the inability to operate in const context),
>>> we can introduce a new setter than works with a const argument and
>>> spares us the need to invoke `Bounded::new` ourselves:
>>> 
>>>      let color = Rgb::default().
>>>              .with_red::<0x10>()
>>>              .with_green::<0x1f>()
>>>              .with_blue::<0x18>()
>>
>> Are we sure that .with_red is a better name than, say, .set_red()?
>>
>> "with" is not so easy to remember, because it is a bit
>> surprising and different, for setting a value.
>>
>> "with" feels like a function call or closure: "sort with
>> qsort", for example. But here we are setting a color
>> component.
>
> `set_foo` implies that the value is mutated in place (and takes `&mut self`).
> `with_foo` implies that value is returned with the specific thing changed. For
> example, `pointer::with_addr`, `Path::with_extension`.
>
> Given the signature in the API I would agree with Yury that `with_` is better.

Indeed, given the signature of the method `with_` is definitely more
idiomatic. But that is also true for the non-const setter, and they
unfortunately cannot share the same name.

Alternative names I can think of for the non-const / const setters:

- `with_foo`/`with_foo_const`, but that's a bit verbose,
- `with_foo`/`const_foo`, but that's not idiomatic either,
- `with_foo_val`/`with_foo`, but that increases the risk of name
  clashes.

... so unless there are better proposals I guess the
`set_foo`/`with_foo` is at least practical, if a bit unconventional.