Generate fully typed rust code from your EdgeDB schema and inline queries.
To install the edgedb_codegen
crate you can use the following command.
cargo add edgedb_codegen
Or directly add the following to your Cargo.toml
file.
edgedb_codegen = "0.2"
Follow the Quickstart Guide to make sure your edgedb instance is running. The macro relies on the running edgedb
instance to parse the output of the provided query string.
When working with edgedb
you often need to write queries and also provide the typed for both the input and output. Your code is only checked at runtime which increases the risk of bugs and errors.
Fortunately, edgedb
has a query language that is typed and can be converted into types and queried for correctness at compile time.
use edgedb_codegen::edgedb_query;
use edgedb_errors::Error;
use edgedb_tokio::create_client;
// Creates a module called `simple` with a function called `query` and structs
// for the `Input` and `Output`.
edgedb_query!(
simple,
"select { hello := \"world\", custom := <str>$custom }"
);
#[tokio::main]
async fn main() -> Result<(), Error> {
let client = create_client().await?;
let input = simple::Input {
custom: String::from("custom"),
};
// For queries the following code can be used.
let output = simple::query(&client, &input).await?;
Ok(())
}
The macro above generates the following code:
pub mod simple {
use ::edgedb_codegen::exports as e;
/// Execute the desired query.
#[cfg(feature = "query")]
pub async fn query(
client: &e::edgedb_tokio::Client,
props: &Input,
) -> core::result::Result<Output, e::edgedb_errors::Error> {
client.query_required_single(QUERY, props).await
}
/// Compose the query as part of a larger transaction.
#[cfg(feature = "query")]
pub async fn transaction(
conn: &mut e::edgedb_tokio::Transaction,
props: &Input,
) -> core::result::Result<Output, e::edgedb_errors::Error> {
conn.query_required_single(QUERY, props).await
}
#[derive(Clone, Debug)]
#[cfg_attr(feature = "builder", derive(e::typed_builder::TypedBuilder))]
#[cfg_attr(feature = "query", derive(e::edgedb_derive::Queryable))]
#[cfg_attr(feature = "serde", derive(e::serde::Serialize, e::serde::Deserialize))]
pub struct Input {
#[cfg_attr(feature = "builder", builder(setter(into)))]
pub custom: String,
}
impl e::edgedb_protocol::query_arg::QueryArgs for Input {
fn encode(
&self,
encoder: &mut e::edgedb_protocol::query_arg::Encoder,
) -> core::result::Result<(), e::edgedb_errors::Error> {
let map = e::edgedb_protocol::named_args! {
"custom" => self.custom.clone(),
};
map.encode(encoder)
}
}
#[derive(Clone, Debug)]
#[cfg_attr(feature = "query", derive(e::edgedb_derive::Queryable))]
#[cfg_attr(feature = "serde", derive(e::serde::Serialize, e::serde::Deserialize))]
pub struct Output {
pub hello: String,
pub custom: String,
}
/// The original query string provided to the macro. Can be reused in your
/// codebase.
pub const QUERY: &str = "select { hello := \"world\", custom := <str>$custom }";
}
Define a query file in the queries
directory of your crate called select_user.edgeql
.
# queries/select_user.edgeql
select User {
name,
bio,
slug,
} filter .slug = <str>$slug;
Then use the edgedb_query
macro to import the query.
use edgedb_codegen::edgedb_query;
use edgedb_errors::Error;
use edgedb_tokio::create_client;
// Creates a module called `select_user` with public functions `transaction` and
// `query` as well as structs for the `Input` and `Output`.
edgedb_query!(select_user);
#[tokio::main]
async fn main() -> Result<(), Error> {
let client = create_client().await?;
// Generated code can be run inside a transaction.
let result = client
.transaction(|mut txn| {
async move {
let input = select_user::Input {
slug: String::from("test"),
};
let output = select_user::transaction(&mut txn, &input).await?;
Ok(output)
}
})
.await?;
Ok(())
}
This crate is still in early development and there are several features that are not yet implemented.
Currently the following types are not supported:
enum
- currently all enums are represented as strings.MultiRange
- The macro will panic if a multirange is used.
Currently all enums are represented as strings.
In order to support full enum generation the edgedb-protocol
crate needs to be updated to use the binary protocol 2.0. In the current 1.0 version the enum descriptors are returned without the name property.
Once this is implemented the macro will be able to generate the correct code.
However end users probably don't want multiple enums for each generated query module as this would break sharing. To get around this, there should be a macro for generating the shared types used by all other.
// lib.rs
use edgedb_codegen::generate_shared_types;
generate_shared_types!(); // exports the shared types to the `edb` module.
These are not currently exported by the edgedb-protocol
so should be added in a PR to the edgedb-protocol
crate, if they are still supported in the new protocol.
Currently everything is hardcoded and the macro is not configurable.
The following configuration options should be added:
- Name of input struct (optional) -
Input
by default. - Name of output struct (optional) -
Output
by default. - Name of query function (optional) -
query
by default. - Name of transaction function (optional) -
transaction
by default. - Default location of queries (optional) -
queries
by default. - Default crate export name for shared types (optional) -
edb
by default. - Default
edgedb
instance (optional) -$EDGEDB_INSTANCE
by default. - Default
edgedb
branch (optional) -$EDGEDB_BRANCH
by default.
Probably these should be read from the Cargo.toml
file and parsed manually to prevent slowdowns from parsing the file.
Currently the macro depends on having a running edgedb instance to parse the query string.
Once an LSP is created for edgedb it would make sense to switch from using string to using inline edgedb queries.
use edgedb_codegen::edgedb_query;
edgedb_query!(
example,
select User {**}
);
Create a edgedb_codegen_cli
crate which supports generating the typed code into rust files rather than inline queries. This is useful for larger projects to prevent constantly compiling the queries on every change / build.
default
— The default feature iswith_all
.with_bigint
— Include thenum-bigint
dependency.with_bigdecimal
— Use thebigdecimal
crate.with_chrono
— Use thechrono
crate for all dates.with_all
(enabled by default) — Include all additional types. This is included by default. Usedefault-features = false
to disable.builder
— Use thetyped-builder
crate to generate the builders for the generatedInput
structs.query
— Turn on thequery
andtransaction
methods and anything that relies onedgedb-tokio
. The reason to separate this feature is to enable usage of this macro in browser environments whereedgedb-tokio
is not feasible.serde
— Enable serde for the generated code.
devenv
is used to provide a reproducible development environment for this project. Follow the getting started instructions.
To automatically load the environment you should install direnv and then load the direnv
.
# The security mechanism didn't allow to load the `.envrc`.
# Since we trust it, let's allow it execution.
direnv allow .
At this point you should see the nix
commands available in your terminal.
Run the following commands to install all the required dependencies.
install:all
This installs all the cargo binaries locally so you don't need to worry about polluting your global namespace.
At this point you must setup the edgedb instance.
db:setup # setup the edgedb instance
Now you can make your changes and run tests.
test:all
build:all
: Build all crates with all features activated.build:docs
: Build documentation site.coverage:all
: Test all files and generate a coverage report for upload to codecov.db:destroy
: Destroy the local database.db:setup
: Setup the local database.db:up
: Watch changes to the local database.fix:all
: Fix all fixable lint issues.fix:clippy
: Fix fixable lint issues raised by rust clippy.fix:format
: Fix formatting for entire project.install:all
: Install all dependencies.install:cargo:bin
: Install cargo binaries locally.lint:all
: Lint all project files.lint:clippy
: Check rust clippy lints.lint:format
: Check all formatting is correct.setup:ci
: Setup the github ci environment.setup:helix
: Setup the helix editor for development.setup:vscode
: Setup the vscode editor for development.test:all
: Test all project files.update:deps
: Update dependencies.
If you have an outdated version of devenv
you can update it by running the following commands. If you have an easier way, please create a PR and I'll update these docs.
nix profile list # find the index of the devenv package
nix profile remove <index>
nix profile install ---accept-flake-config nixpkgs#devenv
To setup recommended configuration for your favorite editor run the following commands.
setup:vscode # Setup vscode with recommended configuration
setup:helix # Setup helix with recommended configuration
Unlicense, see the license file.