What’s New?
- Hive accounts for argument, input field, and variable values when checking for breaking changes – This improvement is fully backwards compatible and will automatically take effect when upgrading your hive client package.
Feature Details
Breaking change detection is complicated. It starts in the graphql-hive client packages, where operations are processed into a list of requested fields and types.
For example, an operation { user(id: "1234") { id } } will produce this list of coordinates:
- Query
- Query.user
- Query.user.id
- User
- User.id
- ID
These coordinates, along with other useful metrics like latency and error status, are sent to Hive’s servers in batches from the client(s). Hive processes and stores this usage data by timestamp and coordinate. These are then aggregated over time to produce what’s seen on Hive’s explorer and insights pages, as well as to detect whether not a specific change to the schema will break clients.
This approach works well, but previously, it only collected from the type definitions used by the operation. But this release adds an additional coordinate for inputs with an indicator if the data is provided by clients, either as a value or variable.
For the above example, an additional coordinate is generated:
- Query.user.id!
The exclamation (!) indicates that this argument was passed a value. This helps distinguish whether or not an input field or argument can be converted from nullable to non-nullable safely.
We are excited about this feature because it provides greater accuracy and convenience for our users, while keeping their schemas safe.
For more information about conditional breaking changes, see the Target Management docs.