flow-immutable-models
This repository contains a codemod script for use with
JSCodeshift that creates model classes backed by Immutable.js data structures based on Flow type interfaces.
Setup with yarn
yarn global add jscodeshift
git clone https://github.com/pbomb/flow-immutable-models.git
or download a zip file
from https://github.com/pbomb/flow-immutable-models/archive/master.zip
- Run
yarn install
in the flow-immutable-models directory - Copy the ImmutableModel.js base class into your repository
Setup with npm
npm install -g jscodeshift
git clone https://github.com/pbomb/flow-immutable-models.git
or download a zip file
from https://github.com/pbomb/flow-immutable-models/archive/master.zip
- Run
npm install
in the flow-immutable-models directory - Copy the ImmutableModel.js base class into your repository
Executing the codemod script
cp /path/to/flow-immutable-models/src/ImmutableModel.js src/models
jscodeshift -t /path/to/flow-immutable-models/src/transform.js <path>
Use the -d
option for a dry-run and use -p
to print the output for comparison. For more information about the jscodeshift CLI options, check out its README.
How it works
This codemod modifies any file that exports Flow type declarations named like *Interface
. For each matching exported Flow type, a model class will be created. If this script is re-run and the model class already exists, it will be processed again and replaced, meaning it is safe to run this script multiple times against the same files.
Class design
Immutable collections are great for simplifying application development by avoid object mutations and enabling performance optimizations such as memoization and reference equality comparisons. A popular immutable collection library is Immutable.js.
One downside to using Immutable.js collections (Immutable.List, Immutable.Map, etc.) is that the objects do not lend themselves to static analysis / typing with tools like Flow or TypeScript. For example, with Flow we often end up typing Maps like Immutable.Map<string, any>
; This means that the map contains unknown keys of type string values can be of any kind. It says nothing about which keys are allowed and what type a value for a given key should be. While there are some ways of providing better typing than this, there are still gaps in how well these objects can be desribed.
This codemod library takes the approach of wrapping an Immutable.Map
with a typed ES6 class definition. As a consumer, you would create files with exported Flow type definitions described as an Object with defined keys and values. Running this codemod against these files creates an ES6 class with getters and setters for each typed property. Each setter function returns a new instance of the class so that you can continue to take advantage of performance optimizations like memoization and reference equality checking since the class instances are immutable.
Examples
In your repo, models/User.js
exists as
import * as Immutable from 'immutable';
import ImmutableModel from './ImmutableModel.js';
export type UserInterface = {
id: number,
name: string,
address: AddressInterface,
};
Each exported type matching *Interface
(in this case UserInterface
) must be described as an object literal. Otherwise, the script will throw an error.
Running jscodeshift -t /path/to/flow-immutable-models/src/transform.js **/models/*.js
will update models/User.js
to be
import * as Immutable from 'immutable';
import ImmutableModel from './ImmutableModel.js';
export type UserInterface = {
id: number,
name: string,
};
export class User extends ImmutableModel {
static fromJS(state: UserInterface): User {
return this.fromImmutable(Immutable.fromJS(state));
}
static fromImmutable(ImmutableModel: Immutable.Map<string, any>): User {
return new User(ImmutableModel);
}
initialize(): Immutable.Map<string, any> {
let state = Immutable.Map();
return state;
}
get id(): number {
return this._state.get('id');
}
setId(id: number): User {
return new User(this._state.set('id'));
}
get name(): string {
return this._state.get('name');
}
setName(name: string): User {
return new User(this._state.set('name'));
}
}
Creating a new user can be done a few different ways. Here are some examples:
const user = User.fromJS({
id: 1,
name: "Floyd O' Phone",
});
const user = User.fromImmutable(
Immutable.Map(
{
id: 1,
name: "Floyd O' Phone",
}
)
);
const user = new User()
.setId(1)
.setName("Floyd O' Phone");
Once you have a User
instance, you can get properties using property accessor syntax. For instance, the user's name can be retrieved with:
const name = user.name;
Setting a property will always return a new instance of User
.
const user1 = User.fromJS({ id: 1, name: "Floyd O' Phone"});
const user2 = user1.setName("leeb");
console.log(user1 === user2);
Nested Objects
If you want to have a nested model class, each nested object should be declared as its own type.
Example:
export type User {
name: string,
};
export type GithubIssue {
author: User,
body: string,
title: string,
}
Then you can interact with the generated models like this:
const issue = GithubIssue.fromJS({
author: {
name: "kittens"
},
title: "Yarn is too fast",
body: "I can no longer sword-fight while installing dependencies: https://xkcd.com/303/"
});
console.log(issue.author.name);
const updatedIssue = issue.setAuthor(issue.author.setName("thejameskyle"));
console.log(updatedIssue.author.name);
Recast Options
Options to recast's printer can be provided
through the printOptions
command line argument
jscodeshift -t transform.js <path> --printOptions='{ "quote":"double" }'
The default options are
{ "quote": "single": "trailingComma": true }