1 unstable release

0.1.0 Feb 26, 2024

#2758 in Parser implementations

Download history 15/week @ 2024-07-19 31/week @ 2024-07-26 35/week @ 2024-08-02 25/week @ 2024-08-09 15/week @ 2024-08-16 35/week @ 2024-08-23 24/week @ 2024-08-30 15/week @ 2024-09-06 19/week @ 2024-09-13 47/week @ 2024-09-20 42/week @ 2024-09-27 42/week @ 2024-10-04 169/week @ 2024-10-11 62/week @ 2024-10-18 10/week @ 2024-10-25 13/week @ 2024-11-01

259 downloads per month
Used in mdbook-tabs

BSD-3-Clause

73KB
1.5K SLoC

A lightweight serde deserializer for strings containing key-value pairs separated by commas, as commonly found in command-line parameters.

Say your program takes a command-line option of the form:

--foo type=bar,active,nb_threads=8

This crate provides a [from_key_values] function that deserializes these key-values into a configuration structure. Since it uses serde, the same configuration structure can also be created from any other supported source (such as a TOML or YAML configuration file) that uses the same keys.

Integration with the argh command-line parser is also provided via the argh_derive feature.

The deserializer supports parsing signed and unsigned integers, booleans, strings (quoted or not), paths, and enums inside a top-level struct. The order in which the fields appear in the string is not important.

Simple example:

use serde_keyvalue::from_key_values;
use serde::Deserialize;

#[derive(Debug, PartialEq, Deserialize)]
struct Config {
    path: String,
    threads: u8,
    active: bool,
}

let config: Config = from_key_values("path=/some/path,threads=16,active=true").unwrap();
assert_eq!(config, Config { path: "/some/path".into(), threads: 16, active: true });

let config: Config = from_key_values("threads=16,active=true,path=/some/path").unwrap();
assert_eq!(config, Config { path: "/some/path".into(), threads: 16, active: true });

As a convenience the name of the first field of a struct can be omitted:

#[derive(Debug, PartialEq, Deserialize)]
struct Config {
    path: String,
    threads: u8,
    active: bool,
}

let config: Config = from_key_values("/some/path,threads=16,active=true").unwrap();
assert_eq!(config, Config { path: "/some/path".into(), threads: 16, active: true });

Fields that are behind an Option can be omitted, in which case they will be None.

#[derive(Debug, PartialEq, Deserialize)]
struct Config {
    path: Option<String>,
    threads: u8,
    active: bool,
}

let config: Config = from_key_values("path=/some/path,threads=16,active=true").unwrap();
assert_eq!(config, Config { path: Some("/some/path".into()), threads: 16, active: true });

let config: Config = from_key_values("threads=16,active=true").unwrap();
assert_eq!(config, Config { path: None, threads: 16, active: true });

Alternatively, the serde default attribute can be used on select fields or on the whole struct to make unspecified fields be assigned their default value. In the following example only the path parameter must be specified.

#[derive(Debug, PartialEq, Deserialize)]
struct Config {
    path: String,
    #[serde(default)]
    threads: u8,
    #[serde(default)]
    active: bool,
}

let config: Config = from_key_values("path=/some/path").unwrap();
assert_eq!(config, Config { path: "/some/path".into(), threads: 0, active: false });

A function providing a default value can also be specified, see the serde documentation for field attributes for details.

Booleans can be true or false, or take no value at all, in which case they will be true. Combined with default values this allows to implement flags very easily:

#[derive(Debug, Default, PartialEq, Deserialize)]
#[serde(default)]
struct Config {
    active: bool,
    delayed: bool,
    pooled: bool,
}

let config: Config = from_key_values("active=true,delayed=false,pooled=true").unwrap();
assert_eq!(config, Config { active: true, delayed: false, pooled: true });

let config: Config = from_key_values("active,pooled").unwrap();
assert_eq!(config, Config { active: true, delayed: false, pooled: true });

Strings can be quoted, which is useful if they need to include a comma or a bracket, which are considered separators for unquoted strings. Quoted strings can also contain escaped characters, where any character after a \ is repeated as-is:

#[derive(Debug, PartialEq, Deserialize)]
struct Config {
    path: String,
}

let config: Config = from_key_values(r#"path="/some/\"strange\"/pa,th""#).unwrap();
assert_eq!(config, Config { path: r#"/some/"strange"/pa,th"#.into() });

Tuples and vectors are allowed and must be specified between [ and ]:

#[derive(Debug, PartialEq, Deserialize)]
struct Layout {
    resolution: (u16, u16),
    scanlines: Vec<u16>,
}

let layout: Layout = from_key_values("resolution=[320,200],scanlines=[0,64,128]").unwrap();
assert_eq!(layout, Layout { resolution: (320, 200), scanlines: vec![0, 64, 128] });

Enums can be directly specified by name. It is recommended to use the rename_all serde container attribute to make them parseable using snake or kebab case representation. Serde's rename and alias field attributes can also be used to provide shorter values:

#[derive(Debug, PartialEq, Deserialize)]
#[serde(rename_all="kebab-case")]
enum Mode {
    Slow,
    Fast,
    #[serde(rename="ludicrous")]
    LudicrousSpeed,
}

#[derive(Deserialize, PartialEq, Debug)]
struct Config {
    mode: Mode,
}

let config: Config = from_key_values("mode=slow").unwrap();
assert_eq!(config, Config { mode: Mode::Slow });

let config: Config = from_key_values("mode=ludicrous").unwrap();
assert_eq!(config, Config { mode: Mode::LudicrousSpeed });

A nice use of enums is along with sets, where it allows to e.g. easily specify flags:

#[derive(Deserialize, PartialEq, Eq, Debug, PartialOrd, Ord)]
#[serde(rename_all = "kebab-case")]
enum Flags {
    Awesome,
    Fluffy,
    Transparent,
}
#[derive(Deserialize, PartialEq, Debug)]
struct TestStruct {
    flags: BTreeSet<Flags>,
}

let res: TestStruct = from_key_values("flags=[awesome,fluffy]").unwrap();
assert_eq!(
    res,
    TestStruct {
        flags: BTreeSet::from([Flags::Awesome, Flags::Fluffy]),
    }
);

Enums taking a single value can use the flatten field attribute in order to be inferred from their variant key directly:

#[derive(Debug, PartialEq, Deserialize)]
#[serde(rename_all="kebab-case")]
enum Mode {
    // Work with a local file.
    File(String),
    // Work with a remote URL.
    Url(String),
}

#[derive(Deserialize, PartialEq, Debug)]
struct Config {
    #[serde(flatten)]
    mode: Mode,
}

let config: Config = from_key_values("file=/some/path").unwrap();
assert_eq!(config, Config { mode: Mode::File("/some/path".into()) });

let config: Config = from_key_values("url=https://www.google.com").unwrap();
assert_eq!(config, Config { mode: Mode::Url("https://www.google.com".into()) });

The flatten attribute can also be used to embed one struct within another one and parse both from the same string:

#[derive(Debug, PartialEq, Deserialize)]
struct BaseConfig {
    enabled: bool,
    num_threads: u8,
}

#[derive(Debug, PartialEq, Deserialize)]
struct Config {
    #[serde(flatten)]
    base: BaseConfig,
    path: String,
}

let config: Config = from_key_values("path=/some/path,enabled,num_threads=16").unwrap();
assert_eq!(
    config,
    Config {
        path: "/some/path".into(),
        base: BaseConfig {
            num_threads: 16,
            enabled: true,
        }
    }
);

If an enum's variants are made of structs, it can take the untagged container attribute to be inferred directly from the fields of the embedded structs:

#[derive(Debug, PartialEq, Deserialize)]
#[serde(untagged)]
enum Mode {
    // Work with a local file.
    File {
        path: String,
        #[serde(default)]
        read_only: bool,
    },
    // Work with a remote URL.
    Remote {
        server: String,
        port: u16,
    }
}

#[derive(Debug, PartialEq, Deserialize)]
struct Config {
    #[serde(flatten)]
    mode: Mode,
}

let config: Config = from_key_values("path=/some/path").unwrap();
assert_eq!(config, Config { mode: Mode::File { path: "/some/path".into(), read_only: false } });

let config: Config = from_key_values("server=google.com,port=80").unwrap();
assert_eq!(config, Config { mode: Mode::Remote { server: "google.com".into(), port: 80 } });

Using this crate, parsing errors and invalid or missing fields are precisely reported:

#[derive(Debug, PartialEq, Deserialize)]
struct Config {
    path: String,
    threads: u8,
    active: bool,
}

let config = from_key_values::<Config>("path=/some/path,active=true").unwrap_err();
assert_eq!(format!("{}", config), "missing field `threads`");

Most of the serde container and field attributes can be applied to your configuration struct. Most useful ones include deny_unknown_fields to report an error if an unknown field is met in the input, and deserialize_with to use a custom deserialization function for a specific field.

Be aware that the use of flatten comes with some severe limitations. Because type information is not available to the deserializer, it will try to determine the type of fields using the input as its sole hint. For instance, any number will be returned as an integer type, and if the parsed structure was actually expecting a number as a string, then an error will occur. Struct enums also cannot be flattened and won't be recognized at all.

For these reasons, it is discouraged to use flatten except when neither the embedding not the flattened structs has a member of string type.

Most of the time, similar functionality can be obtained by implementing a custom deserializer that parses the embedding struct's member and then the flattened struct in a specific order using the key_values::KeyValueDeserializer interface directly.

Another limitation of using flatten that is inherent to serde is that it won't allow deny_unknown_fields to be used in either the embedding or the flattened struct.

Dependencies

~1.2–2.1MB
~43K SLoC