#serde #proc-macro #optionals #maybe-types


Tools to define and convert between types and their corresponding "maybe types"

4 releases

0.2.2 Jun 22, 2022
0.2.1 Jun 22, 2022
0.2.0 Jun 22, 2022
0.1.0 Jun 21, 2022

#1067 in Data structures

Download history 30/week @ 2024-02-19 16/week @ 2024-02-26 10/week @ 2024-03-04 14/week @ 2024-03-11 6/week @ 2024-03-18 87/week @ 2024-04-01

109 downloads per month
Used in 3 crates (via webpack-stats)


457 lines

Empty Type

The EmptyType trait and Container trait work together to create structures with optional members and provides an api convert between the two.

use empty_type::{EmptyType, Empty};

#[empty(deserialize, default)]
struct Data {
    key: String,
    mismatch: usize

const JSON: &str = r#"{ "key": "value", "mismatch": "not a number" }"#; 

fn main() {
    let empty: Empty<Data> = serde_json::from_str(JSON).unwrap();
    let resolved = empty.resolve();
    assert_eq!(resolved.key.as_str(), "value");
    // when the "default" flag is set, even serde errors get resolved
    assert_eq!(resolved.mismatch, 0);

The proc macro creates code roughly equivalent to the following

use empty_type::{EmptyType, Container};

struct Data {
    key: String

struct OptionalData {
   key: Option<String>

impl EmptyType for Data {
    type Container = OptionalData;

impl Container for OptionalData {
#    type Value = Data;
#    fn try_open(&mut self) -> Result<Self::Value, Box<dyn std::error::Error>> {
#        Ok(Data {
#            key: self.key.open()
#        })
#    }

// This allows conversion between the two types 

fn main() {
    let mut empty = Data::new_empty();
    empty.key = Some(String::new());
    // should be the default value
    let resolved = empty.resolve();

Proc Macro

The behavior above is tedious and complicated. The proc_macro EmptyType creates the optional data structures for you with the feature derive enabled


Serde support is provided by the feature flag serde and a helper function deserialize_empty is provided to deserialize empty values

# use empty_type::{EmptyType, deserialize_empty};
# use serde::Deserialize;

struct Data {
    value: String

const JSON: &str = r#" { "value": "data" } "#;

fn use_with_deserializer() {
    let mut de = serde_json::Deserializer::from_str(JSON);
    let empty_value = deserialize_empty::<Data, _>(&mut de).unwrap();
    let full_value = empty_value.resolve();
    assert_eq!(full_value.value.as_str(), "data");

fn use_implicit_deserialization() {
    let value: empty_type::Empty<Data> = serde_json::from_str(JSON).unwrap();
    let full_value = value.resolve();

    assert_eq!(full_value.value.as_str(), "data");

# fn main() {
#   use_with_deserializer();
# }


Container is automatically implemented for Option<T> and bool. This allows container unwraps to propagate up through containers.


A special container types Fallible and Optional provide small variations to the way that types are opened.


Optional is a wrapper around source types that are initially option. Optional roughly represents Some(Option<T>). Opening this container always results in an Option This is distinct from Option<Option<T>> as it's impossible for the wrapping option to be None with the semantics described.

use empty_type::Optional;
struct Data {
    optional_data: Option<String>

struct MaybeData {
    optional_data: Optional<Option<String>>


Fallible is similar to Optional except it requires that the underlying type implement Default. The semantics of fallible are to always return the default value of the underlying Container.

Another important distinction is that Fallible will swallow serde Deserialize errors. Any error in deserialization will result in the default type being emitted.