Deser Project Aims to Address Serde Serialization Limitations in Rust

Motivation for Deser

Serde is a prominent serialization library in the Rust ecosystem, but it presents specific limitations that can lead to developer frustration. The Deser project was initiated to offer an alternative that resolves these issues, providing a more robust and predictable serialization framework.

Arbitrary Precision Number Handling

One limitation in Serde involves the interaction of internally tagged enums with the `serde_json` arbitrary precision feature. When this feature is enabled, Serde's data model lacks a direct representation for arbitrary precision numbers, leading to errors when deserializing specific JSON structures. Deser aims to provide a more consistent approach to handling such numerical types.

Flatten Attribute and JSON Keys

Another issue arises with Serde's `flatten` attribute when deserializing JSON objects containing integer keys. While `serde_json` can convert string keys to integers when directly deserializing a `HashMap<u32, u32>`, this conversion fails when the `HashMap` is part of a flattened structure. The `flatten` attribute buffers the value as a string, preventing the intended integer conversion. Deser seeks to rectify this behavior, ensuring correct deserialization regardless of the `flatten` attribute's presence.

Custom Deserializers for Optional and Collection Types

Serde also exhibits limitations in applying custom deserializer functions to types wrapped in `Option`, `Vec`, or other collections. A single custom deserializer function cannot be directly applied to the inner type of an `Option<T>` or `Vec<T>`. This forces developers to write separate deserializer functions for each wrapper type, increasing code duplication. Deser aims to simplify this process, allowing for more flexible application of custom deserialization logic.