synclane 0.4.0

framework-agnostic RPC API with a smart auto-generated TypeScript client

Welcome to synclane

synclane is a framework-agnostic RPC API with a smart auto-generated TypeScript client.

Idea

The below must be enough to define an API:

class UserParams(pydantic.BaseModel):
    uid: str

class GetUsers(AbstractProcedure):
    def call(self, in_: UserParams, context) -> List[UserDetails]:
        ...

and use an automatically generated frontend TypeScript client:

import { callGetUsers } from "./src/out";

expect(callGetUsers(userParams).$promise).resolves.toEqual(listOfUserDetails);

Benefits

Automated typescript client generation

Of course, it's possible to annotate your API, export an OpenAPI schema and generate a typescript client from it. However it will lack the below nice bits.

Browser Dates done right

Javascript doesn't have a separate date type, so it uses Date for both python's date and datetime.

Hence when you pass 2000-01-01 to a browser in New York, the browser will read it as UTC datetime and then convert it to the local timezone, so it will give you Dec 31, 1991 7:00PM, which is fine if you wanted to work with a particular moment in time, but what if you wanted to display someone's date of birth? That's why lacking date type is a problem.

synclane will see that you wanted to pass python's date to the browser and will automatically prepare it in the browser, so that Jan 1st is preserved in the case above.

Browser friendly types only

synclane raises an exception if you use types, which browser won't be able to understand.

No need to define URLs

Once you name a procedure, e.g. AddUser, you just get callAddUser function in the typescript client. You don't need to define any other identifier like API endpoint url.

Enums

If your procedure in/out types include enums, they will become available in the typescript client.

Installation

pip install synclane

pydantic is the only dependency.

Usage

1. define procedures
2. define RPC instance, its error handling method, register procedures and dump TypeScript client code
3. connect RPC to an API
4. on TypeScript side: import rpcConfig and initialize:
   • rpcConfig.url: url where RPC is listening
   • rpcConfig.initFetch (optional): function, which accepts and can mutate fetch options as needed

Example

Step 1: Define procedures

--8<-- "tests/int_tst/main.py:def_procedures"

Step 2: Define RPC, dump TS

--8<-- "tests/int_tst/main.py:def_rpc"

Step 3.a: Connect to Django

/// tab | async rpc

--8<-- "tests/int_tst/main.py:django_async"

///

/// tab | sync rpc

--8<-- "tests/int_tst/main.py:django_sync"

///

Step 3.b: Connect to FastAPI

/// tab | async rpc

--8<-- "tests/int_tst/main.py:fastapi_async"

///

/// tab | async rpc

--8<-- "tests/int_tst/main.py:fastapi_sync"

///

Step 4: Use autogenerated TS client

--8<-- "tests/int_tst/tests/client.test.ts:imports"
--8<-- "tests/int_tst/tests/client.test.ts:rpc_config"
--8<-- "tests/int_tst/tests/client.test.ts:get_user"