Auto Some-wrapping of Option<T> config values

[Boscop]

E.g. if the Config struct contains a foo: Option<i32>, a value of foo: 42 in the config should be accepted, and converted to Some(42) automatically, instead of failing to deserialize.

[kvark]

I believe this is what the newtype extension does - #72

[Boscop]

@kvark I don't mean newtype wrapping, I meant if my struct has a field #[serde(default)] foo: Option<i32> then currently the RON file has to contains foo: Some(42), even though if I want None, I just omit the field.
With serde_json, writing foo: 42 works, too (becomes Some(42)).

It could be said that's not a good behavior if there is ambiguity in cases like Option<Option<i32>>
but I think it doesn't have to be ambiguous, because if the RON file contains 42 it would unambiguously be Some(Some(42)) and if one wants to express Some(None), it would be written like that in the RON file.
The only ambiguous case is if the RON file contains None, but that can be said to mean None, not Some(None), because if Some(None) is meant, that should be written.
So the disambiguation rule would be to do the least interpretation work:
Deserializing None as None involves less interpretation than deserializing None as Some(None).

So it seems to be non-ambiguous, and it would really improve the usability of handwriting RON files to have auto-Some wrapping :)

[kvark]

Ah sorry, you are totally right, discard my previous comment 😅

[jonasbb]

@Boscop If you are willing to use third-party crates you could try serde_with::rust::unwrap_or_skip. It was written for RON. It only requires some attributes on the optional fields: #[serde(default, skip_serializing_if = "Option::is_none", with = "::serde_with::rust::unwrap_or_skip")]

[Boscop]

@jonasbb Thanks. (Btw, this issue was about deserializing T into Option<T> which is now supported via the implicit_some extension.)
But yes, I also often want the reverse: Skipping None during serialization (and serializing only val instead of Some(val) in that case.
So the point of with = "::serde_with::rust::unwrap_or_skip" instead of just #[serde(default, skip_serializing_if = "Option::is_none")] is that Some(val) will serialize as just val, right?

[jonasbb]

@Boscop The listed combination of attributes works for serialization and deserialization. This allows you to round trip the data. It also allows you to omit the Some(_) part without explicitly listing it as an extension.
Thanks, I didn't know about the extensions yet, I should check how implicit_some interacts with serde_with::rust::unwrap_or_skip.

So the point of with = "::serde_with::rust::unwrap_or_skip" instead of just #[serde(default, skip_serializing_if = "Option::is_none")] is that Some(val) will serialize as just val, right?

Yes, exactly. This then allows you to also serialize these documents.

[Boscop]

It also allows you to omit the Some(_) part without explicitly listing it as an extension.

That's a great advantage, the extension attribute line confuses people who come from JSON.
I wish I could also enable the unwrap_newtypes extension statically (#281) so that it also works when serializing (#284).

[github-actions]

Issue has had no activity in the last 180 days and is going to be closed in 7 days if no further activity occurs