Use Rust enums directly with diesel ORM.
The latest release, 1.1.1, is tested against diesel 1.4 and rustc 1.39.0. It may work with older versions.
Note: The current master branch tracks diesel master, and will not work with diesel 1.x.
toml
[dependencies]
diesel-derive-enum = { version = "1", features = ["..."] } # "postgres", "mysql" or "sqlite"
```rust // define your enum
pub enum MyEnum { Foo, // All variants must be fieldless Bar, BazQuxx, }
// define your table table! { use diesel::types::Integer; use super::MyEnumMapping; mytable { id -> Integer, someenum -> MyEnumMapping, // Generated Diesel type - see below for explanation } }
// define a struct with which to populate/query the table
struct MyRow { id: i32, some_enum: MyEnum, } ```
Postgres: ```sql -- by default the postgres ENUM values correspond to snakecased Rust enum variant names CREATE TYPE myenum AS ENUM ('foo', 'bar', 'baz_quxx');
CREATE TABLE mytable (
id SERIAL PRIMARY KEY,
someenum myenum NOT NULL
);
MySQL:
sql
CREATE TABLE mytable (
id SERIAL PRIMARY KEY,
myenum enum('foo', 'bar', 'bazquxx') NOT NULL -- note: snakecase
);
sqlite:
sql
CREATE TABLE mytable (
id SERIAL PRIMARY KEY,
myenum TEXT CHECK(myenum IN ('foo', 'bar', 'bazquxx')) NOT NULL -- note: snakecase
);
```
Now we can insert and retrieve MyEnum directly:
rust
let data = vec![
MyRow {
id: 1,
some_enum: MyEnum::Foo,
},
MyRow {
id: 2,
some_enum: MyEnum::BazQuxx,
},
];
let connection = PgConnection::establish(/*...*/).unwrap();
let inserted = insert_into(my_table::table)
.values(&data)
.get_results(&connection)
.unwrap();
assert_eq!(data, inserted);
Postgres arrays work too! See this example.
Enums work slightly differently in each of the three databases.
* In Postgres, the user declares an enum as a separate type within a schema, which may then
be used in multiple tables. Internally, an enum value is encoded as an int (four bytes)
and stored inline within a row - a much more efficient representation than a string.
* MySQL is similar except the enum is not declared as a separate type and is 'local' to
it's parent table. It is encoded as either one or two bytes.
* sqlite does not have enums - in fact, it does not really have types;
you can store any kind of data in any column. Instead we emulate static checking by
adding the CHECK command, as per above. This does not give a more compact encoding
but does ensure data integrity. Note that if you somehow retreive some other invalid
text as an enum, diesel will error at the point of deserialization.
Diesel maintains a set of internal types which correspond one-to-one to the types available in various
relational databases. Each internal type in turn maps to some kind of Rust native type.
e.g. Postgres INTEGER maps to diesel::types::Integer maps to i32. Therefore when we create a new enum in Postgres
with CREATE TYPE ..., we must also create a corresponding type in Diesel, then map it to
some native Rust type (our enum). That is the purpose of this crate.
If you are getting compilation errors, it could be that these three types names are not 'in sync'.
By default, the database and Diesel internal types are inferred from the name of the Rust enum.
Specifically, we assume MyEnum corresponds to my_enum in your database schema and MyEnumMapping in Diesel.
These defaults can be overridden with the attributes #[PgType = "..."] and #[DieselType = "..."].
(The PgType annotation has no effect on MySQL or sqlite).
Similarly, by default we assume that the possible ENUM variants are simply the Rust enum variants
translated to snake_case. These can be renamed with the inline annotation #[db_rename = "..."].
See this test for an example of renaming.
You can override the snake_case assumption for the entire enum using the #[DbValueStyle = "..."] attribute. Individual variants can still be renamed using #[db_rename = "..."].
| DbValueStyle | Variant | Value | |:-------------------:|:---------:|:---| | camelCase | BazQuxx | "bazQuxx" | | kebab-case | BazQuxx | "baz-quxx" | | PascalCase | BazQuxx | "BazQuxx" | | SCREAMINGSNAKE | BazQuxx | "BAZQUXX" | | snakecase | BazQuxx | "bazquxx" | | verbatim | Bazquxx | "Bazquxx" |
See this test for an example of changing the output style.
print-schema and infer-schema!The print-schema command (from diesel_cli) attempts to connect to an existing DB and generate
a correct mapping of Postgres columns to Diesel internal types. If a custom ENUM exists in the
database, Diesel will simply assume that the internal mapping type is the ENUM name,
Title-cased (e.g. my_enum -> My_enum). Therefore the derived mapping name must also
be corrected with the DieselType attribute e.g. #[DieselType = "My_enum"].
If you are using a diesel.toml file to generate your schema.rs file, you should be sure to
add an extra import for the enum mapping type, i.e. it should look something like the following:
``` toml
[printschema] file = "src/schema.rs" importtypes = ["diesel::sqltypes::*", "crate::myenum::*"] # <- note the extra import ```
Unfortunately the infer_schema! is not compatible with this crate.
Licensed under either of these: