Am 09.02.26 um 09:45 schrieb Dietmar Maurer:
> Introduce a new CommaSeparatedList<T> wrapper type that provides
> schema-aware serialization and deserialization of comma-separated
> values, similar to PropertyString but designed for list/array types.
> 
> Key components:
> - CommaSeparatedListSchema trait: Provides the static ARRAY_SCHEMA
>   required for (de)serialization (workaround for unstable generic
>   const items in Rust)
> - CommaSeparatedList<T>: A transparent Vec<T> newtype with Deref/
>   DerefMut implementations for ergonomic access
> 
> The wrapper automatically handles conversion between "1,2,3" string
> representation and Vec<T> while validating against the element schema.
> 
> Signed-off-by: Dietmar Maurer <[email protected]>
> ---
> 
> Changes in v2:
> 
> - use description from ARRAY_SCHEMA
> - add test for that
> 
> 
> 
>  proxmox-schema/src/comma_separated_list.rs | 173 +++++++++++++++++++++
>  proxmox-schema/src/lib.rs                  |   1 +
>  2 files changed, 174 insertions(+)
>  create mode 100644 proxmox-schema/src/comma_separated_list.rs
> 
> diff --git a/proxmox-schema/src/comma_separated_list.rs 
> b/proxmox-schema/src/comma_separated_list.rs
> new file mode 100644
> index 00000000..381b17c5
> --- /dev/null
> +++ b/proxmox-schema/src/comma_separated_list.rs
> @@ -0,0 +1,173 @@
Would be nice to have a module-level rust doc for new modules with an usage 
example.
As those doc examples are run on testing/build, if not excluded explicitly, 
they could
also provide some (extra) coverage here.

E.g., from a quick check the cases for empty list serializing to "" (or?) and a 
list with spaces after the comma are not tested as of now.

> ...

> +impl<T: FromStr + Display> CommaSeparatedList<T> {

Is there any reason for limiting this to FromStr + Display ?
The implementation does not depend on it and if one would have a type with
Serialize + CommaSeparatedListSchema implemented, they couldn't use this 
constructor.

If there's a reason behind doing this, it'd be nice to have a short comment here
to describe why the type bounds are more limiting than necessary.

> +    pub fn new(inner: Vec<T>) -> Self {
> +        Self(inner)
> +    }
> +
> +    pub fn into_inner(self) -> Vec<T> {
> +        self.0
> +    }
> +}
> +

> ...

> diff --git a/proxmox-schema/src/lib.rs b/proxmox-schema/src/lib.rs
> index 1647e8a9..fd773a84 100644
> --- a/proxmox-schema/src/lib.rs
> +++ b/proxmox-schema/src/lib.rs
> @@ -22,6 +22,7 @@ pub mod de;
>  pub mod format;
>  pub mod ser;
>  
> +pub mod comma_separated_list;

This makes the main type accessible under 
"proxmox_schema::comma_separated_list::CommaSeparatedList"
which is rather redundant and IMO has not much benefits over re-exporting 
CommaSeparatedList
directly to allow accessing it under proxmox_schema::CommaSeparatedList

But no hard feelings, we got this "problem" quite a few times already.



Reply via email to