v1.57.0
Important
If you have enabled Distributed query plan caching, updates to the query planner in this release will result in query plan caches being re-generated rather than re-used. On account of this, you should anticipate additional cache regeneration cost when updating between these versions while the new query plans come into service.
🚀 Features
Remove legacy schema introspection (PR #6139)
Schema introspection in the router now runs natively without JavaScript. We have high confidence that the new native implementation returns responses that match the previous Javascript implementation, based on differential testing: fuzzing arbitrary queries against a large schema, and testing a corpus of customer schemas against a comprehensive query.
Changes to the router's YAML configuration:
- The
experimental_introspection_mode
key has been removed, with thenew
mode as the only behavior in this release. - The
supergraph.query_planning.legacy_introspection_caching
key is removed, with the behavior in this release now similar to what wasfalse
: introspection responses are not part of the query plan cache but instead in a separate, small in-memory—only cache.
When using the above deprecated configuration options, the router's automatic configuration migration will ensure that existing configuration continue to work until the next major version of the router. To simplify major upgrades, we recommend reviewing incremental updates to your YAML configuration by comparing the output of ./router config upgrade --config path/to/config.yaml
with your existing configuration.
By @SimonSapin in #6139
Support new request_context
selector for telemetry (PR #6160)
The router supports a new request_context
selector for telemetry that enables access to the supergraph schema ID.
You can configure the context to access the supergraph schema ID at the router service level:
telemetry:
instrumentation:
events:
router:
my.request_event:
message: "my request event message"
level: info
on: request
attributes:
schema.id:
request_context: "apollo::supergraph_schema_id" # The key containing the supergraph schema id
You can use the selector in any service at any stage. While this example applies to events
attributes, the selector can also be used on spans and instruments.
Support reading and setting port
on request URIs using Rhai (Issue #5437)
Custom Rhai scripts in the router now support the request.uri.port
and request.subgraph.uri.port
functions for reading and setting URI ports. These functions enable you to update the full URI for subgraph fetches. For example:
fn subgraph_service(service, subgraph){
service.map_request(|request|{
log_info(``);
if request.subgraph.uri.port == {} {
log_info("Port is not explicitly set");
}
request.subgraph.uri.host = "api.apollographql.com";
request.subgraph.uri.path = "/api/graphql";
request.subgraph.uri.port = 1234;
log_info(``);
});
}
🐛 Fixes
Fix various edge cases for __typename
field (PR #6009)
The router now correctly handles the __typename
field used on operation root types, even when the subgraph's root type has a name that differs from the supergraph's root type.
For example, given a query like this:
{
...RootFragment
}
fragment RootFragment on Query {
__typename
me {
name
}
}
Even if the subgraph's root type returns a __typename
that differs from Query
, the router will still use Query
as the value of the __typename
field.
This change also includes fixes for other edge cases related to the handling of __typename
fields. For a detailed technical description of the edge cases that were fixed, please see this description.
By @IvanGoncharov in #6009
Support uri
and method
properties on router "request" objects in Rhai (PR #6147)
The router now supports accessing request.uri
and request.method
properties from custom Rhai scripts. Previously, when trying to access request.uri
and request.method
on a router request in Rhai, the router would return error messages stating the properties were undefined.
An example Rhai script using these properties:
fn router_service(service) {
let router_request_callback = Fn("router_request_callback");
service.map_request(router_request_callback);
}
fn router_request_callback (request) {
log_info(`Router Request... Host: , Path: `);
}
By @andrewmcgivery in #6114
Cost calculation for subgraph requests with named fragments (PR #6162)
In some cases where subgraph GraphQL operations contain named fragments and abstract types, demand control used the wrong type for cost calculation, and could reject valid operations. Now, the correct type is used.
This fixes errors of the form:
Attempted to look up a field on type MyInterface, but the field does not exist
By @goto-bus-stop in #6162
Federation v2.9.3 (PR #6161)
This release updates to Federation v2.9.3, with query planner fixes:
- Fixes a query planning bug where operation variables for a subgraph query wouldn't match what's used in that query.
- Fixes a query planning bug where directives applied to
__typename
may be omitted in the subgraph query. - Fixes a query planning inefficiency where some redundant subgraph queries were not removed.
- Fixes a query planning inefficiency where some redundant inline fragments in
@key
/@requires
selection sets were not optimized away. - Fixes a query planning inefficiency where unnecessary subgraph jumps were being added when using
@context
/@fromContext
.
By @sachindshinde in #6161