dataloader-codegen is an opinionated JavaScript library for automagically generating DataLoaders over a set of resources (e.g. HTTP endpoints), with a predictable interface, and maintains type safety.
Read more about the motivation behind this library in our recent blog post: https://engineeringblog.yelp.com/2020/04/open-sourcing-dataloader-codegen.html
Features:
- 🚚 Supports Batched + Non Batched Resources
- ✨ Predictable DataLoader Interfaces
- 🐛 Error Handling
- 🔒 Type Safety (Flow)
- 🔧 Resource Middleware
$ yarn add --dev dataloader-codegen
See: https://engineeringblog.yelp.com/2020/04/open-sourcing-dataloader-codegen.html
We believe the DataLoader layer should be (mostly) transparent when implementing a GraphQL server over a set of existing resources (e.g. HTTP API Endpoints).
When fetching data, GraphQL resolver authors should think in terms of the underlying resources that they're already familiar with, not an invented set of human defined DataLoaders.
With dataloader-codegen, we build a 1:1 mapping of resources to DataLoaders:
This makes it super easy to find the DataLoaders you want - there will be exactly 1 DataLoader available per resource, with a predictable name and interface.
This means reduced risk of making unnecessary HTTP requests.
If there were (accidentally!) multiple DataLoaders created for a single endpoint, we potentially lose out on batched requests to that resource.
By keeping the mapping of one DataLoader per resource, we reduce that risk and can make a more efficient set of HTTP requests to the underlying resource.
-
Create
dataloader-config.yaml
to describe the shape and behaviour of your resources. (See the docs for detailed info.)Example
resources: getPeople: docsLink: https://swapi.dev/documentation#people isBatchResource: true batchKey: people_ids newKey: person_id getPlanets: docsLink: https://swapi.dev/documentation#planets isBatchResource: true batchKey: planet_ids newKey: planet_id ...
(Can be arbitrarily nested. See the swapi example for an example.)
-
Call
dataloader-codegen
and pass in your config file:$ dataloader-codegen --config swapi.dataloader-config.yaml --output __codegen__/swapi-loaders.js
See
--help
for more options. -
Import the generated loaders and use them in your resolver methods:
import getLoaders from './__codegen__/swapi-loaders'; // StarWarsAPI is a clientlib containing fetch calls to swapi.dev // getLoaders is the function that dataloader-codegen generates for us const swapiLoaders = getLoaders(StarWarsAPI); class Planet { constructor(id) { this.id = id; } async diameter() { const { diameter } = await swapiLoaders.getPlanets.load({ planet_id: this.id }); return diameter; } }
Check out the swapi example to see a working example of this.
The DataLoader .load
interface
accepts a single key and returns a single value. For batch resources, we'll need to transform the DataLoader interface accordingly.
Example
Consider the following resource that returns information about users:
const getUserInfo = (args: {
user_ids: Array<number>,
locale: string,
include_slow_fields?: boolean,
}): Promise<Array<UserInfo>> => fetch('/userInfo', args);
This is a batch resource that accepts a list of users (user_ids
) and returns a list of corresponding user objects (Array<UserInfo>
).
For the DataLoader version of this, we'll want to instead ask for a single user object at a time. This means we need to transform the interface in the following ways:
-
Call
.load
with the same arguments, but switch "user_ids" to "user_id". -
Return a single
UserInfo
object from.load
, instead of an array ofUserInfo
objects.
We can control this by specifying batchKey
and newKey
in the config to
describe the relevant argument in the resource and DataLoader respectively.
The config for our getUserInfo
would therefore look like this:
resources:
getUserInfo:
isBatchResource: true
batchKey: user_ids
newKey: user_id
See the full docs for more information on how to configure resources.
Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.
Please make sure to update tests as appropriate.
See PUBLISH.md