If you want to read more about data loaders in general, you can head over to Facebook's GitHub repository.
Every data fetching technology suffers the n+1 problem. The difference between GraphQL and e.g. REST is, that the n+1 problem occurs on the server, rather than on the client. The clear benefit is, that we only have to deal with this problem once on the server, rather than on every client.
To depict the issue that data loaders solve in this context, let assume we have this schema:
type Query { person(id: ID): Person}
type Person { id: ID name: String friends: [Person]}
The above schema allows to fetch a person by its internal identifier and each person has a list of friends that is represented by a list of persons.
A query against the above schema could look like the following:
{ a: person(id: "a") { name }
b: person(id: "b") { name }}
The above request fetches two persons in one go without the need to call the backend twice. The problem with the GraphQL backend is that field resolvers are atomic and do not have any knowledge about the query as a whole. So, a field resolver does not know that it will be called multiple times in parallel to fetch similar or equal data from the same data source.
The idea of a dataloader is to batch these two requests into one call to the database.
Let's look at some code to understand what data loaders are doing. First, let's have a look at how we would write our field resolver without data loaders:
public async Task<Person> GetPerson(string id, [Service]IPersonRepository repository){ return await repository.GetPersonById(id);}
The above example would result in two calls to the person repository that would then fetch the persons one by one from our data source.
If you think that through you see that each GraphQL request would cause multiple requests to our data source resulting in sluggish performance and unnecessary round-trips to our data source.
This means that we reduced the round-trips from our client to our server with GraphQL but still have the round-trips between the data sources and the service layer.
With data loaders we can now centralize the data fetching and reduce the number of round trips to our data source.
Instead of fetching the data from the repository directly, we fetch the data from the data loader. The data loader batches all the requests together into one request to the database.
// This is one way of implementing a data loader. You will find the different ways of declaring// data loaders further down the page.public class PersonBatchDataLoader : BatchDataLoader<string, Person>{ private readonly IPersonRepository _repository;
public PersonBatchDataLoader( IPersonRepository repository, IBatchScheduler batchScheduler, DataLoaderOptions<string>? options = null) : base(batchScheduler, options) { _repository = repository; }
protected override async Task<IReadOnlyDictionary<string, Person>> LoadBatchAsync( IReadOnlyList<string> keys, CancellationToken cancellationToken) { // instead of fetching one person, we fetch multiple persons var persons = await _repository.GetPersonByIds(keys); return persons.ToDictionary(x => x.Id); }}
public class Query{ public async Task<Person> GetPerson( string id, PersonBatchDataLoader dataLoader) => await dataLoader.LoadAsync(id);}
Execution
With a data loader, you can fetch entities with a key. These are the two generics you have in the class data loaders:
public class BatchDataLoader<TId, TEntity>
TId
is used as an identifier of TEntity
. TId
is the type of the values you put into LoadAsync
.
The execution engine of Hot Chocolate tries to batch as much as possible. It executes resolvers until the queue is empty and then triggers the data loader to resolve the data for the waiting resolvers.
Data Consistency
Dataloader do not only batch calls to the database, they also cache the database response. A data loader guarantees data consistency in a single request. If you load an entity with a data loader in your request more than once, it is given that these two entities are equivalent.
Data loaders do not fetch an entity if there is already an entity with the requested key in the cache.
Types of Data loaders
In Hot Chocolate you can declare data loaders in two different ways. You can separate the data loading concern into separate classes or you can use a delegate in the resolver to define data loaders on the fly. Below you will find the different types of data loaders with examples for class and delegate definition.
Batch DataLoader
One - To - One, usually used for fields like
personById
or one to one relations
The batch data loader collects requests for entities and sends them as a batch request to the data source. Moreover, the data loader caches the retrieved entries within a request.
The batch data loader gets the keys as IReadOnlyList<TKey>
and returns an IReadOnlyDictionary<TKey, TValue>
.
Class
public class PersonBatchDataLoader : BatchDataLoader<string, Person>{ private readonly IPersonRepository _repository;
public PersonBatchDataLoader( IPersonRepository repository, IBatchScheduler batchScheduler, DataLoaderOptions<string>? options = null) : base(batchScheduler, options) { _repository = repository; }
protected override async Task<IReadOnlyDictionary<string, Person>> LoadBatchAsync( IReadOnlyList<string> keys, CancellationToken cancellationToken) { // instead of fetching one person, we fetch multiple persons var persons = await _repository.GetPersonByIds(keys); return persons.ToDictionary(x => x.Id); }}
public class Query{ public async Task<Person> GetPerson( string id, PersonBatchDataLoader dataLoader) => await dataLoader.LoadAsync(id);}
Delegate
public Task<Person> GetPerson( string id, IResolverContext context, [Service] IPersonRepository repository){ return context.BatchDataLoader<string, Person>( async (keys, ct) => { var result = await repository.GetPersonByIds(keys); return result.ToDictionary(x => x.Id); }) .LoadAsync(id);}
An example with the Batch DataLoader can be found here.
Group DataLoader
One - To - Many, usually used for fields like
personsByLastName
or one to many relations
The group data loader is also a batch data loader but instead of returning one entity per key, it returns multiple entities per key. As with the batch data loader retrieved collections are cached within a request.
The group data loader gets the keys as IReadOnlyList<TKey>
and returns an ILookup<TKey, TValue>
.
Class
public class PersonsByLastNameDataloader : GroupedDataLoader<string, Person>{ private readonly IPersonRepository _repository;
public PersonsByLastNameDataloader( IPersonRepository repository, IBatchScheduler batchScheduler, DataLoaderOptions<string>? options = null) : base(batchScheduler, options) { _repository = repository; }
protected override async Task<ILookup<string, Person>> LoadGroupedBatchAsync( IReadOnlyList<string> names, CancellationToken cancellationToken) { var persons = await _repository.GetPersonsByLastName(names); return persons.ToLookup(x => x.LastName); }}
public class Query{ public async Task<IEnumerable<Person>> GetPersonByLastName( string lastName, PersonsByLastNameDataloader dataLoader) => await dataLoader.LoadAsync(lastName);}
Delegate
public Task<IEnumerable<Person>> GetPersonByLastName( string lastName, IResolverContext context, [Service]IPersonRepository repository){ return context.GroupDataLoader<string, Person>( async (keys, ct) => { var result = await repository.GetPersonsByLastName(keys); return result.ToLookup(t => t.LastName); }) .LoadAsync(lastName);}
Cache DataLoader
No batching, just caching. This data loader is used rarely. You most likely want to use the batch data loader.
The cache data loader is the easiest to implement since there is no batching involved. You can just use the initial GetPersonById
method. We do not get the benefits of batching with this one, but if in a query graph the same entity is resolved twice we will load it only once from the data source.
public Task<Person> GetPerson(string id, IResolverContext context, [Service]IPersonRepository repository){ return context.CacheDataLoader<string, Person>("personById", keys => repository.GetPersonById(keys)).LoadAsync(id);}
Stacked DataLoader Calls
This is more like an edge case that is supported than a certain type of data loader. Sometimes we have more complex resolvers that might first fetch data from one data loader and use that to fetch data from the next.
public Task<IEnumerable<Customer>> GetCustomers( string personId, PersonByIdDataLoader personByIdDataLoader, CustomerByIdsDataLoader customerByIdsDataloader){ Person person = await personByIdDataLoader.LoadAsync(personId); return await customerByIdsDataloader.LoadAsync(person.CustomerIds);}