Changelog

Highlights

GraphQL Type Generator in Actions

While creating a new Action, now you can use a type generator on the Console to easily create GraphQL types from JSON sample requests

Aggregate functions in where-clauses

We now support aggregation functions in where-clauses, initially requested in #1498, see docs.

This allows queries such as the following to fetch the names of students that have an average grade above a certain value across their science classes:

query {
  students(_where: {
   classes_aggregate: {
     average: {
       arguments: "grade",
       filter: {department: {_eq: "science"}},
       predicate: { _gt: 10 }
     }
  })
 {
   name
 }

Bug fixes and improvements

Server

• Aggregation Predicates now generate graphql types names which are less prone to conflicts. The entire feature can now be disabled via the experimental feature flag hide_aggregation_predicates.
• add IF EXISTS while dropping the constraint in source migrations (2 -> 3)
• pg_dump has been upgraded in order to support PostgreSQL 15.
• fix batch_size behaviour for auto event trigger log cleanup (Enterprise edition only)
• Use only accepted requests for calculating rate limiting (Enterprise edition only)
• When conflicting type definitions are encountered, show their definitions in the error message
• Remove spaces before colons in error messages and descriptions
• BigQuery: Strings in suitable notation are now permitted for all numerical input fields.
• Fix type name generation in the GraphQL schema for aggregation predicate functions to avoid conflicts
• Fix swapping of the the idle_timeout and max_connections fields for MySQL connection pool settings while applying metadata.
• Replacing the metadata will now reload all affected sources. Previously, the GraphQL Engine would re-use the cached data from the sources which could lead to a stale schema.
• Add origin of errors in error messages of streaming subscription to help find source of errors
• Extend the Config API to report whether the Prometheus metrics API has been enabled. (Enterprise edition only)
• Add hide_update_many_fields experiment feature flag which allows GraphQL Engine to be started without including _updates fields. This is useful for resolving errors where the new fields conflict with existing table names. (fix #8844)
• Avoid exporting the backend_only flag for delete permissions if it is set to false while exporting metadata to make it consistent with insert and update permissions.
• Support count-star aggregates in CockroachDB.
• Fix bug that didn't stream any data when the cursor column of a streaming subscription had a custom identifier set. (fix #8985)

Console

• Fixes failures/indefinite loading during operation inspection in monitoring tab. (Enterprise edition only)
• Add the functionality to customize the GraphQL field name for the update_many operation
• Fix dropping of existing configured response transforms while updating an action
• Show event trigger invocation logs on the console if only event logs are cleaned.
• Fix auto-cleanup frequent frequencies dropdown on event triggers modify page. (Enterprise edition only)
• Fix local object relationship deletion in new relationships page (fix #9024)
• Fix inserting of JSON object for Citus and CockroachDB (fix #9015)
• Fix browse rows page filters for BigQuery and add new operators (in, not in, like, not like) (fix #9023)
• Add Try it button on the table pages to quickly try GraphQL operations on the table using the API explorer.
• Fix export to CSV on the table browse rows page when special characters are present in the data. (fix #7537)

CLI

• Correctly export query_collections.yaml when query_collections object in metadata JSON contains at least one definition.queries as an empty array. (fixes #8787)