On Wed Mar 26, 2025 at 11:01 PM CET, Rob Herring wrote:
>>
>> + let prop = dev
>> + .property_read::<bool>(c_str!("test,bool-prop"))
>> + .required()?;
>
> The 'required' is kind of odd for boolean properties. They are never
> required as not present is the only way to to get false.
Agreed. I can think of a few alternatives:
* Make the trait `Property` more flexible to allow each implementor to specify
what its output type for the `read` function is, via an associated type.
I really don't like this idea, because overly generic APIs can mess with type
inference and become less ergonomic because of it.
* Use `propert_present` instead. That doesn't perfectly express the intention,
because it doesn't warn if the property is present but has a type other than
bool.
* Add an additional inherent method `property_read_bool`, which returns a plain
`bool` instead of `PropertyGuard<bool>`. Then there will be three slightly
different ways to read a bool: `property_present`, `property_read_bool` and
`property_read::<bool>`. Maybe that's confusing.
* Add `property_read_bool` and remove `impl Property for bool`. That would avoid
confusion between `property_read_bool` and `property_read::<bool>`, only the
former would work.
