Re: [PATCH net-next v3 2/8] rust: time: Introduce Delta type

[Miguel Ojeda <miguel.ojeda.sandonis@gmail.com>]

On Fri, Oct 18, 2024 at 6:55 PM Andrew Lunn <andrew@lunn.ch> wrote:
>
> Did you see my other comment, about not actually using these helpers?
> I _guess_ it was not used because it does not in fact round up. So at
> least for this patchset, the helpers are useless. Should we be adding
> useless helpers? Or should we be adding useful helpers? Maybe these
> helpers need a different name, and do actually round up?

Yeah, I saw that -- if I understand you correctly, you were asking why
`as_nanos()` is used and not `as_secs()` directly (did you mean
`as_micros()`?) by adding rounding on `Delta`'s `as_*()` methods.

So my point here was that a method with a name like `as_*()` has
nothing to do with rounding, so I wouldn't add rounding here (it would
be misleading).

Now, we can of course have rounding methods in `Delta` for convenience
if that is something commonly needed by `Delta`'s consumers like
`fsleep()`, that is fine, but those would need to be called
differently, e.g. `to_micros_ceil`: `to_` since it is not "free"
(following e.g. `to_radians`) and + `_ceil` to follow `div_ceil` from
the `int_roundings` feature (and shorter than something like
`_rounded_up`).

In other words, I think you see these small `as_*()` functions as
"helpers" to do something else, and thus why you say we should provide
those that we need (only) and make them do what we need (the
rounding). That is fair.

However, another way of viewing these is as typical conversion methods
of `Delta`, i.e. the very basic interface for a type like this. Thus
Tomonori implemented the very basic interface, and since there was no
rounding, then he added it in `fsleep()`.

I agree having rounding methods here could be useful, but I am
ambivalent as to whether following the "no unused code" rule that far
as to remove very basic APIs for a type like this.

Cheers,
Miguel