Skip to content

Latest commit

 

History

History
762 lines (593 loc) · 28.6 KB

README.md

File metadata and controls

762 lines (593 loc) · 28.6 KB
Raw

MIT License All Contributors PRs Welcome

Star on GitHub Watch on GitHub Discord

Build Status Coverage version

GraphQL Client

graphql/client.dart is a GraphQL client for dart modeled on the apollo client, and is currently the most popular GraphQL client for dart. It is co-developed alongside graphql_flutter on github, where you can find more in-depth examples. We also have a lively community alongside the rest of the GraphQL Dart community on discord.

As of v4, it is built on foundational libraries from the gql-dart project, including gql, gql_link, and normalize. We also depend on hive for persistence via HiveStore.

Useful API Docs:

Installation

First, depend on this package:

dependencies:
  graphql: ^4.0.0-beta

And then import it inside your dart code:

import 'package:graphql/client.dart';

Migration Guide

Find the migration from version 3 to version 4 here.

Basic Usage

To connect to a GraphQL Server, we first need to create a GraphQLClient. A GraphQLClient requires both a cache and a link to be initialized.

In our example below, we will be using the Github Public API. we are going to use HttpLink which we will concatenate with AuthLink so as to attach our github access token. For the cache, we are going to use GraphQLCache.

// ...

final _httpLink = HttpLink(
  'https://api.github.com/graphql',
);

final _authLink = AuthLink(
  getToken: () async => 'Bearer $YOUR_PERSONAL_ACCESS_TOKEN',
);

Link _link = _authLink.concat(_httpLink);

/// subscriptions must be split otherwise `HttpLink` will. swallow them
if (websocketEndpoint != null){
  final _wsLink = WebSocketLink(websockeEndpoint);
  _link = Link.split((request) => request.isSubscription, _wsLink, _link);
}

final GraphQLClient client = GraphQLClient(
  /// **NOTE** The default store is the InMemoryStore, which does NOT persist to disk
  cache: GraphQLCache(),
  link: _link,
);

// ...

Persistence

In v4, GraphQLCache is decoupled from persistence, which is managed (or not) by its store argument. We provide a HiveStore for easily using hive boxes as storage, which requires a few changes to the above:

NB: This is different in graphql_flutter, which provides await initHiveForFlutter() for initialization in main

GraphQL getClient() async {
  ...
  /// initialize Hive and wrap the default box in a HiveStore
  final store = await HiveStore.open(path: 'my/cache/path');
  return GraphQLClient(
      /// pass the store to the cache for persistence
      cache: GraphQLCache(store: store),
      link: _link,
  );
}

Once you have initialized a client, you can run queries and mutations.

Options

All graphql methods accept a corresponding *Options object for configuring behavior. These options all include policies with which to override defaults, an optimisticResult for snappy client-side interactions, gql_exec Context with which to make requests, and of course a document to be requested.

Internally they are converted to gql_exec Requests with .asRequest for execution via links, and thus can also be used with the direct cache access api.

Query

Creating a query is as simple as creating a multiline string:

const String readRepositories = r'''
  query ReadRepositories($nRepositories: Int!) {
    viewer {
      repositories(last: $nRepositories) {
        nodes {
          __typename
          id
          name
          viewerHasStarred
        }
      }
    }
  }
''';

Then create a QueryOptions object:

NB: for document - Use our built-in help function - gql(query) to convert your document string to ASTs document.

In our case, we need to pass nRepositories variable and the document name is readRepositories.

const int nRepositories = 50;

final QueryOptions options = QueryOptions(
  document: gql(readRepositories),
  variables: <String, dynamic>{
    'nRepositories': nRepositories,
  },
);

And finally you can send the query to the server and await the response:

// ...

final QueryResult result = await client.query(options);

if (result.hasException) {
  print(result.exception.toString());
}

final List<dynamic> repositories =
    result.data['viewer']['repositories']['nodes'] as List<dynamic>;

// ...

Mutations

Creating a mutation is similar to creating a query, with a small difference. First, start with a multiline string:

const String addStar = r'''
  mutation AddStar($starrableId: ID!) {
    action: addStar(input: {starrableId: $starrableId}) {
      starrable {
        viewerHasStarred
      }
    }
  }
''';

Then instead of the QueryOptions, for mutations we will MutationOptions, which is where we pass our mutation and id of the repository we are starring.

// ...

final MutationOptions options = MutationOptions(
  document: gql(addStar),
  variables: <String, dynamic>{
    'starrableId': repositoryID,
  },
);

// ...

And finally you can send the mutation to the server and await the response:

// ...

final QueryResult result = await client.mutate(options);

if (result.hasException) {
  print(result.exception.toString());
  return;
}

final bool isStarred =
    result.data['action']['starrable']['viewerHasStarred'] as bool;

if (isStarred) {
  print('Thanks for your star!');
  return;
}

// ...

GraphQL Upload

gql_http_link provides support for the GraphQL Upload spec as proposed at https://github.com/jaydenseric/graphql-multipart-request-spec

mutation($files: [Upload!]!) {
  multipleUpload(files: $files) {
    id
    filename
    mimetype
    path
  }
}
import "package:http/http.dart" show Multipartfile;

// ...

final myFile = MultipartFile.fromString(
  "",
  "just plain text",
  filename: "sample_upload.txt",
  contentType: MediaType("text", "plain"),
);

final result = await graphQLClient.mutate(
  MutationOptions(
    document: gql(uploadMutation),
    variables: {
      'files': [myFile],
    },
  )
);

Subscriptions

To use subscriptions, a subscription-consuming link must be split from your HttpLink or other terminating link route:

link = Link.split((request) => request.isSubscription, websocketLink, link);

Then you can subscribe to any subscriptions provided by your server schema:

final subscriptionDocument = gql(
  r'''
    subscription reviewAdded {
      reviewAdded {
        stars, commentary, episode
      }
    }
  ''',
);
// graphql/client.dart usage
subscription = client.subscribe(
  SubscriptionOptions(
    document: subscriptionDocument
  ),
);
subscription.listen(reactToAddedReview)

client.watchQuery and ObservableQuery

client.watchQuery can be used to execute both queries and mutations, then reactively listen to changes to the underlying data in the cache.

final observableQuery = client.watchQuery(
  WatchQueryOptions(
    fetchResults: true,
    document: gql(
      r'''
      query HeroForEpisode($ep: Episode!) {
        hero(episode: $ep) {
          name
        }
      }
      ''',
    ),
    variables: {'ep': 'NEWHOPE'},
  ),
);

/// Listen to the stream of results. This will include:
/// * `options.optimisitcResult` if passed
/// * The result from the server (if `options.fetchPolicy` includes networking)
/// * rebroadcast results from edits to the cache
observableQuery.stream.listen((QueryResult result) {
  if (!result.isLoading && result.data != null) {
    if (result.hasException) {
      print(result.exception);
      return;
    }
    if (result.isLoading) {
      print('loading');
      return;
    }
    doSomethingWithMyQueryResult(myCustomParser(result.data));
  }
});
// ... cleanup:
observableQuery.close();

ObservableQuery is a bit of a kitchen sink for reactive operation logic – consider looking at the API docs if you'd like to develop a deeper understanding.

client.watchMutation

The default CacheRereadPolicy of client.watchQuery merges optimistic data from the cache into the result on every related cache change. This is great for queries, but an undesirable default for mutations, as their results should not change due to subsequent mutations.

While eventually we would like to decouple mutation and query logic, for now we have client.watchMutation (used in the Mutation widget of graphql_flutter) which has the default policy CacheRereadPolicy.ignoreAll. Otherwise, its behavior is exactly the same. It still takes WatchQueryOptions and returns ObservableQuery, and both methods can take either mutation or query documents. The watchMutation method should be thought of as a stop-gap.

See Rebroadcasting for more details.

NB: watchQuery, watchMutation, and ObservableQuery currently don't have a nice APIs for update onCompleted and onError callbacks, but you can have a look at how graphql_flutter registers them through onData in Mutation.runMutation.

Normalization

The GraphQLCache automatically normalizes data from the server, and heavily leverages the normalize library. Data IDs are pulled from each selection set and used as keys in the cache. The default approach is roughly:

String dataIdFromObject(Map<String, Object> data) {
  final typename = data['__typename'];
  if (typename == null) return null;

  final id = data['id'] ?? data['_id'];
  return id == null ? null : '$typename:$id';
}

To disable cache normalization entirely, you could pass (data) => null. If you only cared about nodeId, you could pass (data) => data['nodeId'].

Here's a more detailed example where the system involved contains versioned entities you don't want to clobber:

String customDataIdFromObject(Map<String, Object> data) {
    final typeName = data['__typename'];
    final entityId = data['entityId'];
    final version = data['version'];
    if (typeName == null || entityId == null || version == null){
      return null;
    }
    return '${typeName}/${entityId}/${version}';
}

Direct Cache Access API

The GraphQLCache leverages normalize to give us a fairly apollo-ish direct cache access API, which is also available on GraphQLClient. This means we can do local state management in a similar fashion as well.

The cache access methods are available on any cache proxy, which includes the GraphQLCache the OptimisticProxy passed to update in the graphql_flutter Mutation widget, and the client itself.

NB counter-intuitively, you likely never want to use use direct cache access methods directly on the cache, as they will not be rebroadcast automatically.
Prefer client.writeQuery and client.writeFragment to those on the client.cache for automatic rebroadcasting

In addition to this overview, a complete and well-commented rundown of can be found in the GraphQLDataProxy API docs.

Request, readQuery, and writeQuery

The query-based direct cache access methods readQuery and writeQuery leverage gql_exec Requests used internally in the link system. These can be retrieved from options.asRequest available on all *Options objects, or constructed manually:

const int nRepositories = 50;

final QueryOptions options = QueryOptions(
  document: gql(readRepositories),
  variables: {
    'nRepositories': nRepositories,
  },
);

var queryRequest = Request(
  operation: Operation(
    document: gql(readRepositories),
  ),
  variables: {
    'nRepositories': nRepositories,
  },
);

/// experimental convenience api
queryRequest = Operation(document: gql(readRepositories)).asRequest(
  variables: {
    'nRepositories': nRepositories,
  },
);

print(queryRequest == options.asRequest);

final data = client.readQuery(queryRequest);
client.writeQuery(queryRequest, data);

The cache access methods are available on any cache proxy, which includes the GraphQLCache the OptimisticProxy passed to update in the graphql_flutter Mutation widget, and the client itself.

NB counter-intuitively, you likely never want to use use direct cache access methods on the cache cache.readQuery(queryRequest); client.readQuery(queryRequest); //

FragmentRequest, readFragment, and writeFragment

FragmentRequest has almost the same api as Request, but is provided directly from graphql for consistency. It is used to access readFragment and writeFragment. The main differences are that they cannot be retreived from options, and that FragmentRequests require idFields to find their cooresponding entities:

final fragmentDoc = gql(
  r'''
    fragment mySmallSubset on MyType {
      myField,
      someNewField
    }
  ''',
);

var fragmentRequest = FragmentRequest(
  fragment: Fragment(
    document: fragmentDoc,
  ),
  idFields: {'__typename': 'MyType', 'id': 1},
);

/// same as experimental convenience api
fragmentRequest = Fragment(document: fragmentDoc).asRequest(
  idFields: {'__typename': 'MyType', 'id': 1},
);

final data = client.readFragment(fragmentRequest);
client.writeFragment(fragmentRequest, data);

NB You likely want to call the cache access API from your client for automatic broadcasting support.

Other Cache Considerations

Write strictness and partialDataPolicy

As of #754 we can now enforce strict structural constraints on data written to the cache. This means that if the client receives structurally invalid data from the network or on client.writeQuery, it will throw an exception.

By default, optimistic data is excluded from these constraints for ease of use via PartialDataCachePolicy.acceptForOptimisticData, as it is easy to miss __typename, etc. This behavior is configurable via GraphQLCache.partialDataPolicy, which can be set to accept for no constraints or reject for full constraints.

Possible cache write exceptions

At link execution time, one of the following exceptions can be thrown:

  • CacheMisconfigurationException if the structure seems like it should write properly, and is perhaps failing due to a typePolicy
  • UnexpectedResponseStructureException if the server response looks malformed.
  • MismatchedDataStructureException in the event of a malformed optimistic result (and PartialDataCachePolicy.reject).
  • CacheMissException if write succeeds but readQuery then returns null (though data will not be overwritten)

Policies

Policies are used to configure various aspects of a request process, and can be set on any *Options object:

// override policies for a single query
client.query(QueryOptions(
  // return result from network and save to cache.
  fetchPolicy: FetchPolicy.networkOnly,
  // ignore all GraphQL errors.
  errorPolicy: ErrorPolicy.ignore,
  // ignore cache data.
  cacheRereadPolicy: CacheRereadPolicy.ignore,
  // ... 
));

Defaults can also be overridden via defaultPolices on the client itself:

GraphQLClient(
 defaultPolicies: DefaultPolicies(
    // make watched mutations behave like watched queries.
    watchMutation: Policies(
      FetchPolicy.cacheAndNetwork,
      ErrorPolicy.none,
      CacheRereadPolicy.mergeOptimistic,
    ),
  ),
  // ... 
)

FetchPolicy: determines where the client may return a result from, and whether that result will be saved to the cache.
Possible options:

  • cacheFirst: return result from cache. Only fetch from network if cached result is not available.
  • cacheAndNetwork: return result from cache first (if it exists), then return network result once it's available.
  • cacheOnly: return result from cache if available, fail otherwise.
  • noCache: return result from network, fail if network call doesn't succeed, don't save to cache.
  • networkOnly: return result from network, fail if network call doesn't succeed, save to cache.

ErrorPolicy: determines the level of events for errors in the execution result.
Possible options:

  • none (default): Any GraphQL Errors are treated the same as network errors and any data is ignored from the response.
  • ignore: Ignore allows you to read any data that is returned alongside GraphQL Errors, but doesn't save the errors or report them to your UI.
  • all: Using the all policy is the best way to notify your users of potential issues while still showing as much data as possible from your server. It saves both data and errors into the Apollo Cache so your UI can use them.

CacheRereadPolicy determines whether and how cache data will be merged into the final QueryResult.data before it is returned. Possible options:

  • mergeOptimistic: Merge relevant optimistic data from the cache before returning.
  • ignoreOptimistic: Ignore optimistic data, but still allow for non-optimistic cache rebroadcasts if applicable.
  • ignoreAll: Ignore all cache data besides the result, and never rebroadcast the result, even if the underlying cache data changes.

Rebroadcasting

Rebroadcasting behavior only applies to watchMutation and watchQuery, which both return an ObservableQuery. There is no rebroadcasting option for subscriptions, because it would be indistiguishable from the previous event in the stream.

Rebroadcasting is enabled unless either FetchPolicy.noCache or CacheRereadPolicy.ignoreAll are set, and whether it considers optimistic results is controlled by the specific CacheRereadPolicy.

Exceptions

If there were problems encountered during a query or mutation, the QueryResult will have an OperationException in the exception field:

/// Container for both [graphqlErrors] returned from the server
/// and any [linkException] that caused a failure.
class OperationException implements Exception {
  /// Any graphql errors returned from the operation
  List<GraphQLError> graphqlErrors = [];

  /// Errors encountered during execution such as network or cache errors
  LinkException linkException;
}

Example usage:

if (result.hasException) {
  if (result.exception.linkException is NetworkException) {
    // handle network issues, maybe
  }
  return Text(result.exception.toString())
}

Links

graphql and graphql_flutter now use the gql_link system, re-exporting gql_http_link, gql_error_link, gql_dedupe_link, and the api from gql_link, as well as our own custom WebSocketLink and AuthLink.

This makes all link development coordinated across the ecosystem, so that we can leverage existing links like gql_dio_link, and all link-based clients benefit from new link development (such as ferry).

Composing Links

NB: WebSocketLink and other "terminating links" must be used with split when there are multiple terminating links.

The gql_link systm has a well-specified routing system: link diagram

a rundown of the composition api:

// kitchen sink:
Link.from([
  // common links run before every request
  DedupeLink(), // dedupe requests
  ErrorLink(onException: reportClientException),
]).split( // split terminating links, or they will break
  (request) => request.isSubscription,
  MyCustomSubscriptionAuthLink().concat(
    WebSocketLink(mySubscriptionEndpoint),
  ), // MyCustomSubscriptionAuthLink is only applied to subscriptions
  AuthLink(getToken: httpAuthenticator).concat(
    HttpLink(myAppEndpoint),
  )
);
// adding links after here would be pointless, as they would never be accessed

/// both `Link.from` and `link.concat` can be used to chain links:
final Link _link = _authLink.concat(_httpLink);
final Link _link = Link.from([_authLink, _httpLink]);

/// `Link.split` and `link.split` route requests to the left or right based on some condition
/// for instance, if you do `authLink.concat(httpLink).concat(websocketLink)`,
/// `websocketLink` won't see any `subscriptions`
link = Link.split((request) => request.isSubscription, websocketLink, link);

When combining links, it isimportant to note that:

  • Terminating links like HttpLink and WebsocketLink must come at the end of a route, and will not call links following them.
  • Link order is very important. In HttpLink(myEndpoint).concat(AuthLink(getToken: authenticate)), the AuthLink will never be called.

AWS AppSync Support

Cognito Pools

To use with an AppSync GraphQL API that is authorized with AWS Cognito User Pools, simply pass the JWT token for your Cognito user session in to the AuthLink:

// Where `session` is a CognitorUserSession
// from amazon_cognito_identity_dart_2
final token = session.getAccessToken().getJwtToken();

final AuthLink authLink = AuthLink(
  getToken: () => token,
);

See more: Issue #209

Other Authorization Types

API key, IAM, and Federated provider authorization could be accomplished through custom links, but it is not known to be supported. Anyone wanting to implement this can reference AWS' JS SDK AuthLink implementation.

Parsing ASTs at build-time

All document arguments are DocumentNodes from gql/ast. We supply a gql helper for parsing, them, but you can also parse documents at build-time use ast_builder from package:gql_code_gen:

dev_dependencies:
  gql_code_gen: ^0.1.5

add_star.graphql:

mutation AddStar($starrableId: ID!) {
  action: addStar(input: { starrableId: $starrableId }) {
    starrable {
      viewerHasStarred
    }
  }
}
import 'package:gql/add_star.ast.g.dart' as add_star;

// ...

final MutationOptions options = MutationOptions(
  document: add_star.document,
  variables: <String, dynamic>{
    'starrableId': repositoryID,
  },
);

// ...

PersistedQueriesLink (experimental) ⚠️ OUT OF SERVICE ⚠️

NOTE: There is a PR for migrating the v3 PersistedQueriesLink, and it works, but requires more consideration. It will be fixed before v4 stable is published

To improve performance you can make use of a concept introduced by apollo called Automatic persisted queries (or short "APQ") to send smaller requests and even enabled CDN caching for your GraphQL API.

ATTENTION: This also requires you to have a GraphQL server that supports APQ, like Apollo's GraphQL Server and will only work for queries (but not for mutations or subscriptions).

You can than use it simply by prepending a PersistedQueriesLink to your normal HttpLink:

final PersistedQueriesLink _apqLink = PersistedQueriesLink(
  // To enable GET queries for the first load to allow for CDN caching
  useGETForHashedQueries: true,
);

final HttpLink _httpLink = HttpLink(
  'https://api.url/graphql',
);

final Link _link = _apqLink.concat(_httpLink);