Flag
This library aims to offer a best-in-class interface for working with feature flags in TypeScript-based React applications.
npm install flag
Getting Started
flag
works by creating bindings at runtime so that context providers, components, and hooks are all strictly typed together. createFlags
builds these bindings without requiring an data.
import { createFlags } from "flag";
export type MyFlags = {
features: {
useMyCoolNewThing: boolean;
};
config: {
apiUrl: string;
};
cool: number;
dude: number;
coolAndDude: number;
largeCoolAndDude: boolean;
};
export const { FlagBackendProvider, Flag, useFlag } = createFlags<MyFlags>();
React Bindings
FlagBackendProvider
Returned as part of createFlags<T>()
.
This React component provides a Backend<T>
(see below) as a data source for Flag
and useFlag
.
Props | Type | Required | Description |
---|
backend | Types.Backend<T> | true | The data source for flags |
children | ReactNode | true | React children |
import React from "react";
import ReactDOM from "react-dom";
import { NullBackend } from "flag";
import { App } from "./app";
import { FlagBackendProvider } from "./flags";
const backend = new NullBackend();
const root = ReactDOM.createRoot(document.querySelector("#app"));
root.render(
<FlagBackendProvider backend={backend}>
<App />
</FlagBackendProvider>
);
useFlag
Returned as part of createFlags<T>()
.
A hook to fetch a single flag. Requires a valid key path and a default value. The key path must terminate at a string, boolean or number and the default value must be of the same type that it terminates. Forcing a default to be provided will minimize the change of a runtime error occurring.
Args | Type | Required | Description |
---|
keyPath | Types.KeyPath<T> | Types.KeyPathString<T> | true | A valid key path of T to a string, boolean or number |
defaultValue | GetValueFromKeyPath<T, KP> | true | A fallback in case it is not available in the Backend<T> |
import { useFlag } from "./flags";
const MyComponent = () => {
const apiUrl = useFlag(["config", "apiUrl"], "https://example.com");
const apiUrl2 = useFlag("config.apiUrl", "https://example.com");
return <div>The API url is "{apiUrl}"</div>;
};
Flag
Returned as part of createFlags<T>()
.
Renders a some UI based on whether a flag is false
or not. (It's a glorified if statement 😬).
Args | Type | Required | Description |
---|
keyPath | Types.KeyPath<T> | Types.KeyPathString<T> | true | A valid key path of T to a string, boolean or number |
defaultValue | GetValueFromKeyPath<T, KP> | true | A fallback in case it is not available in the Backend<T> |
render | (flags: T) => ReactNode | true | Function that returns a ReactNode |
fallback | () => ReactNode | false | Function that returns a ReactNode |
<Flag
name="features.useMyCoolNewThing"
defaultValue={false}
render={() => <div>Rendered on truthy</div>}
fallback={() => <div>Rendered on falsy</div>}
/>
Backends
FlagBackendProvider
requires that you pass a Backend<T>
which is responsible for retreiving flags to your application.
flag
bundles with several useful backends, but you can also roll your own.
StaticBackend<T>
Accepts a JSON object that matches the partial shape of your flags. It can be nested, but shouldn't use arrays.
import React from "react";
import ReactDOM from "react-dom";
import { StaticBackend } from "flag";
import { FlagBackendProvider, MyFlags } from "./flags";
import { App } from "./app";
const backend = new StaticBackend<MyFlags>({
features: {
useMyCoolNewThing: false,
},
config: {
apiUrl: "https://example.com",
},
cool: 100,
dude: 200,
coolAndDude: 300,
largeCoolAndDude: 600,
});
const root = ReactDOM.createRoot(document.querySelector("#app"));
root.render(
<FlagBackendProvider backend={backend}>
<App />
</FlagBackendProvider>
);
If a partial object is provided and a requested key is missing, it will fallback to the provided default. That is,
const backend = new StaticBackend<MyFlags>({
features: {},
});
const SomeScreen = () => {
const newThing = useFlag("features.useMyCoolNewThing", true);
};
ComputedBackend<T>
Similar to StaticBackend
but if a function is used as a value, it will pass in the object T
as an
argument. (Yes, this means you can end up with a stack overflow if you're not careful.) Useful when you need to have composite flags outside of the React render loop.
import React from "react";
import ReactDOM from "react-dom";
import { ComputedBackend } from "flag";
import { FlagBackendProvider, MyFlags } from "./flags";
import { App } from "./app";
const backend = new ComputedBackend<MyFlags>({
features: {
useMyCoolNewThing: false,
},
config: {
apiUrl: "https://example.com",
},
cool: 100,
dude: 200,
coolAndDude: (flags) => flags.cool + flags.dude,
largeCoolAndDude: (flags) => flags.coolAndDude * 2,
});
const root = ReactDOM.createRoot(document.querySelector("#app"));
root.render(
<FlagBackendProvider backend={backend}>
<App />
</FlagBackendProvider>
);
AlwaysBackend
Given a partial mapping of { boolean: boolean; string: string; number: number; }
will always yield the mapping value for a given type. If a type is missing from the mapping, it will fallback to the default value given to useFlag
. Useful for testing.
import React from "react";
import ReactDOM from "react-dom";
import { AlwaysBackend } from "flag";
import { FlagBackendProvider } from "./flags";
import { App } from "./app";
const backend = new AlwaysBackend({
boolean: false,
string: "some string",
number: 1000,
});
const root = ReactDOM.createRoot(document.querySelector("#app"));
root.render(
<FlagBackendProvider backend={backend}>
<App />
</FlagBackendProvider>
);
NullBackend
A backend that always returns the default value. Useful for testing.
import React from "react";
import ReactDOM from "react-dom";
import { NullBackend } from "flag";
import { FlagBackendProvider } from "./flags";
import { App } from "./app";
const backend = new NullBackend();
const root = ReactDOM.createRoot(document.querySelector("#app"));
root.render(
<FlagBackendProvider backend={backend}>
<App />
</FlagBackendProvider>
);
Rolling your own with AbstractBackend<T>
You can roll your own backend by extending a class off of AbstractBackend<T>
. You need only implement the getSnapshot()
(and optionally getServerSnapshot()
) method.
import { AbstractBackend, Types } from "flag";
export class MyBackend<F> extends AbstractBackend<F> {
getSnapshot<KP extends Types.KeyPath<F>, T extends Types.GetValueFromKeyPath<F, KP>>(keyPath: KP, defaultValue: T): T {
return defaultValue;
}
override getServerSnapshot<KP extends Types.KeyPath<F>, T extends Types.GetValueFromKeyPath<F, KP>>(keyPath: KP, defaultValue: T): T {
return defaultValue;
}
}
If you want your backend to work with React Suspense, you can create an async ref object by calling this.createAsyncRef()
and using it in getSnapshot()
.
import { AbstractBackend, Types } from "flag";
export class MyBackend<F> extends AbstractBackend<F> {
#data: Types.AsyncMutableRefObject<T>;
constructor() {
this.#data = this.createAsyncRef();
fetch("/api-with-data")
.then((res) => res.json())
.then((data) => {
this.#data.current = data;
});
}
getSnapshot<KP extends Types.KeyPath<F>, T extends Types.GetValueFromKeyPath<F, KP>>(keyPath: KP, defaultValue: T): T {
const data = this.#data.current;
return someGetterFn(data, keyPath);
}
}
If your backend is asynchronous and you do not want to use suspense, you can imperatively call this.notify()
in order to tell React to re-render.
import { AbstractBackend, Types } from "flag";
export class MyBackend<F> extends AbstractBackend<F> {
#data: T | null = null;
constructor() {
this.#data = this.createAsyncRef();
fetch("/api-with-data")
.then((res) => res.json())
.then((data) => {
this.#data = data;
this.notify();
});
}
getSnapshot<KP extends Types.KeyPath<F>, T extends Types.GetValueFromKeyPath<F, KP>>(keyPath: KP, defaultValue: T): T {
if (this.#data === null) {
return defaultValue;
}
const data = this.#data;
return someGetterFn(data, keyPath);
}
}
Setting NODE_ENV
While in development, you should be sure to set process.env.NODE_ENV
to "development"
for useful warnings when possible. Tool kits like Remix, Next and CRA do this automatically for you.
License
MPL-2.0