apollographql/federation @apollo/gateway@2.4.0

on GitHub

Minor Changes

• This change introduces a configurable query plan cache. This option allows (#2385)
  developers to provide their own query plan cache like so:

  new ApolloGateway({
    queryPlannerConfig: {
      cache: new MyCustomQueryPlanCache(),
    },
  });
  

  The current default implementation is effectively as follows:

  import { InMemoryLRUCache } from "@apollo/utils.keyvaluecache";
  
  const cache = new InMemoryLRUCache<string>({
    maxSize: Math.pow(2, 20) * 30,
    sizeCalculation<T>(obj: T): number {
      return Buffer.byteLength(JSON.stringify(obj), "utf8");
    },
  });
  

  TypeScript users should implement the QueryPlanCache type which is now
  exported by @apollo/query-planner:

  import { QueryPlanCache } from '@apollo/query-planner';
  
  class MyCustomQueryPlanCache implements QueryPlanCache {
    // ...
  }
  

• Adds debug/testing query planner options (debug.bypassPlannerForSingleSubgraph) to bypass the query planning (#2441)
  process for federated supergraph having only a single subgraph. The option is disabled by default, is not recommended
  for production, and is not supported (it may be removed later). It is meant for debugging/testing purposes.

Patch Changes

• Refactor the internal implementation of selection sets used by the query planner to decrease the code complexity and (#2387)
  improve query plan generation performance in many cases.

• Optimises query plan generation for parts of queries that can statically be known to not cross across subgraphs (#2449)

• Updated dependencies [260c357c, 7bc0f8e8, d4426ff9, a9385bdb, 1a555d98, ade7ceb8, 09382e74, cab383b2]:

  • @apollo/composition@2.4.0
  • @apollo/federation-internals@2.4.0
  • @apollo/query-planner@2.4.0