py-ts-interfaces

Python to TypeScript Interfaces

What is this?

This library provides utilities that convert Python dataclasses with type annotations to a TypeScript interface and serializes them to a file.

Installation

python --version  # requires 3.7+
pip install py-ts-interfaces

Motivation

In web applications where Python is used in the backend and TypeScript is used in the frontend, it is often the case that the client will make calls to the backend to request some data with some specific pre-defined "shape". On the client-side, an interface for this data is usually defined and if the Python backend authors use typechecking, like with mypy, the project authors may be typing the JSON response values as well.

This results in a duplication of code. If the shape changes in the backend, the related interface must also be reflect its changes in the frontend. At best, this is annoying to maintain. At worst, over time the interfaces may diverge and cause bugs.

This library aims to have a single source of truth that describes the shape of the payload between the backend and the frontend.

Usage

In Python, py-ts-interfaces exposes a new class object called Interface. By subclassing this object, you identify to the also-packaged script that you want it to be serialized to an interface file.

1. First, hook up your dataclasses:

# views.py
from dataclasses import dataclass
from py_ts_interfaces import Interface

@dataclass
class MyComponentProps(Interface):
    name: str
    show: bool
    value: float

@dataclass
class WillNotGetPickedUp:  # this doesn't subclass Interface, so it won't be included
    name: str
    value: float

2. In your shell, run the included command and pass in the name of the file or directory you want to use. By default it will output to a file in your directory called interface.ts

$ py-ts-interfaces views.py
Created interface.ts!

You may also use the following arguments:

• -o, --output [filepath]: where the file will be saved. default is interface.ts.
• -a, --append: by default each run will overwrite the output file. this flag allows only appends. Be warned, duplicate interfaces are not tested.
• -e, --export: whether the interface definitions should be prepended with export; private python classes will be omitted.

3. The resulting file will look like this:

// interface.ts
interface MyComponentProps {
    name: string;
    show: boolean;
    value: number;
}

Why @dataclass?

Dataclasses were introduced in Python 3.7 and they are great. Some alternatives that I have seen other codebases using are NamedTuple and TypedDict. All of these objects attempt to do the same thing: group together pieces of data that belong close together like a struct.

However, dataclass won out over the other two for the following reasons:

1. dataclasses are built-in to Python. As of writing, NamedTuple is also built-in to the typing module, but TypedDict is still considered experimental.
2. dataclasses cannot be declared and defined inline like you can do with NamedTuple and TypedDict, e.g., NamedTuple can be defined using class inheritance like class MyNamedTuple(NamedTuple): ..., but also like MyNamedTuple = NamedTuple('MyNamedTuple', [('name', str), ('id', int)]). This is a good thing. Dataclasses require you to use a class style declaration, which not only looks closer to a TypeScript interface declaration, but it avoids the complex metaclass machinery that NamedTuples and TypedDicts use to gain all its features. Since this library uses the AST and static analysis of the code to determine what data to serialize, this makes the choice a no-brainer.
3. dataclasses can be made to be immutable (mostly) by setting frozen=True. This library does not require it but in later versions we may provide a partialed dataclass decorator that guarantees immutability.
4. Because we avoid the metaclass machinery of NamedTuples and TypedDicts, it opens up the possibility of writing custom classes that allows mypy to typecheck it one way, but gives the AST parser some clues in order to generate TypeScript types that cannot easily be expressed in Python.

Why define the types in Python instead of TypeScript?

TypeScript is significantly more mature for typing syntax than Python. Generally speaking, you can express any type that Python can do in TypeScript, but not vice versa.

So defining the types in Python guarantee that you can also express the whole interface in both languages.

Supported Type Mappings

Please note that usage of T U and V in the table below represent stand-ins for actual types. They do not represent actually using generic typed variables.

Python	Typescript
None	null
str	string
int	number
float	number
complex	number
bool	boolean
List	Array<any>
Tuple	[any]
Dict	Record<any, any>
List[T]	Array[T]
Tuple[T, U]	[T, U]
Dict[T, U]	Record<T, U>
Optional[T]	T | null
Union[T, U, V]	T | U | V

Planned Supported Mappings

• String literals
• Undefined type
• isNaN type
• ReadOnly types
• Excess Properties

Unsupported/Rejected Mappings

The primary purpose of this library is to help type, first and foremost, data moving back and forth from client to server. Many of these features, whether they be specific to TypeScript or Python, would be overkill to support.

• void
• callables/functions
• enums
• Dates, datetime, dates, times (send these over as strings and convert them to richer objects on the client)
• extends
• generics, TypeVars
• intersection types
• mapped types
• conditional types
• classes

Contributing

Interested in contributing? You're awesome! It's not much, but here's some notes to get you started CONTRIBUTING.md.

Author

Christopher Sabater Cordero