graphql-engine/v3/changelog.md

Changelog

Unreleased

Added

Fixed

Changed

v2024.11.05

Added

JSONAPI alpha release

Adds a new set of endpoints at /v1/rest that follow the JSONAPI specification.

An OpenAPI schema for a given role can be accessed at v1/rest/__schema.

Currently, every model that a given role is able to access is exposed at GET v1/rest/subgraph/model. In further releases models will be explicitly configured and exposed via metadata to match the GraphQL schema.

Select the fields you receive with ?fields[model]=fieldname,anotherfield.

Limit the number of results with ?page[limit]=10

Offset the results with ?page[offset]=5

Order the results with ?sort[model]=fieldname,-anotherfield. Default is sorting in ascending order, adding - at the start of the field name makes the ordering descending instead.

This feature is still very much alpha and in active development, all feedback gratefully received.

v2024.10.30

Added

Order by Nested Fields

Add support for ordering by nested fields.

Example query:

query MyQuery {
  InstitutionMany(order_by: { location: { city: Asc } }) {
    id
    location {
      city
      campuses
    }
  }
}

This will order by the value of the nested field city within the location JSONB column.

• New metadata item OrderByExpression

• New metadata item Model version 2

Fixed

Changed

v2024.10.25

Fixed

• ModelPermissions and CommandPermissions can now correctly preset predicate-typed arguments that are nullable. Previously, trying to do so resulted in a build error.

v2024.10.23

Added

• Add a check to disallow defining boolean expression of array fields with scalar boolean type while resolving the boolean expression

Fixed

• Improve performance of metadata builds

• When the CompatibilityConfig date is set to 2024-10-16 or newer, session variables returned by webhooks, set in noAuth config in AuthConfig or set in JWT claims are now correctly allowed to be full JSON values, not just JSON strings. This fixes the bug where you were incorrectly required to JSON-encode your JSON value inside a string. For example, you were previously incorrectly required to return session variables like this { "X-Hasura-AllowedUserIds": "[1,2,3]" }, but now you can correctly return them like this: { "X-Hasura-AllowedUserIds": [1,2,3] }.

• The warning about AuthConfig v1 being deprecated was only being displayed in the engine's stdout logs and not as a build warning. This has been corrected.

v2024.10.21

Added

• Support array values in session variables

Fixed

Changed

v2024.10.14

Added

• Added contexts to more MBS errors: when a model refers to a collection that doesn't exist, the path to the offending reference will be reported.

Fixed

• Fix local docker-compose.yaml file so that running docker compose up builds the engine and serves it along with a sample schema using ndc-postgres and a postgres database.

• Subgraph builds that have relationships to other external subgraphs can now be run locally and no longer fail with missing subgraph errors. Subgraph builds are marked with a new OpenDD flag and when these builds are run by the engine relationships to unknown subgraphs are automatically pruned.

• Aggregate queries now support __typename introspection fields.

Changed

• metadata-build-service POST endpoints now accept zstd (preferred) or gzip -encoded request bodies

• The --partial-supergraph command-line argument and PARTIAL_SUPERGRAPH environment variable have been removed. Builds now contain an OpenDD flag that indicates if they are subgraph builds and should be run as such.

v2024.10.02

Added

Metadata build error contexts

Contexts are being added to errors raised during the build process to allow users to locate the source of the issue more quickly. These contexts will be surfaced in the Build Server API responses. The first example and test bed for developing the scaffolding is the error raised when a model refers to a nonexistent data connector. This error will now also contain the path to the offending data connector name.

Pre-response Plugin

Engine now supports calling a HTTP webhook in the pre-response execution step. This can be used to add some post execution functionalities to the DDN, such as sending the response to a logging service, sending notifications for specific requests like mutations, etc.

The following is an example of the OpenDD metadata for the pre-response plugin:

kind: LifecyclePluginHook
version: v1
definition:
  name: logging
  url:
    value: http://localhost:5001/log
  pre: response
  config:
    request:
      headers:
        additional:
          hasura-m-auth:
            value: "your-strong-m-auth-key"
      session: {}
      rawRequest:
        query: {}
        variables: {}
      rawResponse: {}

Similar to the pre-parse plugin, the pre-response plugin's request can be customized using the LifecyclePluginHook metadata object. Currently we support the following customizations:

• adding/removing session information
• adding new headers
• forwarding specific headers
• adding/removing graphql query and variables
• adding/removing response

Fixed

• Fix poor performance of process_response for large and deeply-nested results
• Fixed issue in partial supergraph builds where a BooleanExpressionType that referenced a relationship that targeted an unknown subgraph would incorrectly produce an error rather than ignoring the relationship
• Fixed double string escaping when forwarding headers to a data connector

Changed

• Making args non-compulsory for models where all arguments have presets.

Previously, if a model had arguments specified that were all provided by presets, then we would require them to pass an empty args: {} argument:

query MyQuery
  ActorsByMovieMany(args: {}) {
    actor_id
    movie_id
    name
  }
}

This change loosens the restriction, so now the following query is valid too:

query MyQuery
  ActorsByMovieMany {
    actor_id
    movie_id
    name
  }
}

• OpenTelemetry service name set to ddn-engine to avoid confusion with graphql-engine.

• Builds can no longer contain two commands with the same root field name. Previously, one of the two commands would be chosen arbitrarily as the exposed root field. Now, this raises a build-time error.

v2024.09.23

Fixed

• Disallow defining custom scalar types with names that conflict with built-in types, such as String or Int.

• Fixed bug where relationships defined on a boolean expression would not take the target subgraph into account.

• Propagate deprecation status to boolean expression relationship fields.

v2024.09.16

Fixed

• Raise a warning when nested array comparisons are used without the necessary data connector capability. A new OpenDD flag require_nested_array_filtering_capability can be used to promote this warning to an error.

• Disallow recursive types in SQL table column types.

• Previously, if you had AggregateExpressions that were configured to be used in GraphQL, or Models configured for aggregates in GraphQL, but you did not set the appropriate configuration in GraphqlConfig.definition.query.aggregates, the build would fail with an error. This has been relaxed so that the build now succeeds, but warnings are raised instead. However, the aggregates will not appear in your GraphQL API until the GraphqlConfig is updated. This allows you to add AggregateExpressions and configure your Model but update your GraphqlConfig separately, which is useful if they are in separate repositories.

• A build error is now raised if an AggregateExpression specifies an aggregatableField that has field arguments. This is an unsupported scenario and previously would have allowed invalid queries that omitted the required field arguments. These queries may have failed with errors at query time.

• Add a missing typecheck of ValueExpression while resolving model predicates.

v2024.09.05

Added

• SQL endpoint can utilize uniqueness constraints

Fixed

• Fix the name and description of the span resolving relationship predicates in the Engine.

Changed

v2024.09.02

Added

• Enhanced handling of relationships in predicates
• Filter nested arrays
• Order by nested fields
• A new GraphQL config flag require_valid_ndc_v01_version to promote warnings about NDC version as errors.

Enhanced Handling of Relationships in Predicates

Improved support for using relationships in boolean expressions even when the data connector lacks the relation_comparisons capability. This update introduces two strategies for handling relationship predicates:

• Data Connector Pushdown: When the source and target connectors are the same and the target connector supports relationship comparisons, predicates are pushed down to the NDC (Data Connector) for more efficient processing. This strategy optimizes query execution by leveraging the data connector’s capabilities.

• Engine-Based Resolution: When the data connector does not support relationship comparisons or when dealing with relationships targeting models from other data connectors (remote relationships), predicates are resolved internally within the engine. This approach involves querying the target model’s field values and constructing the necessary comparison expressions.

This enhancement updates the GraphQL schema's boolean expression input types by introducing relationship predicates. The feature is gated by a compatibility date to ensure backward compatibility. To enable it, set the date to 2024-09-03 or later in your DDN project's globals/compatibility-config.hml file.

Filter Nested Arrays

If institution is a big JSON document, and staff is an array of objects inside it, we can now filter institutions based on matches that exist within that array.

query MyQuery {
  where_does_john_hughes_work: InstitutionMany(
    where: { staff: { last_name: { _eq: "Hughes" } } }
  ) {
    id
    location {
      city
      campuses
    }
  }

This query would return us details of Chalmers University of Technology, where John Hughes is a member of staff.

Fixed

• Stack overflow error on startup. Even if the (experimental) SQL feature was turned off, engine would try to build a SQL catalog on startup. Now it will build an empty catalog.

Changed

[v2024.08.22]

Added

Pre-parse Engine Plugins

Add support for pre-parse engine plugins. Engine now supports calling a HTTP webhook in pre-parse execution step. This can be used to add a bunch of functionalities to the DDN, such as an allow list.

The following is an example of the OpenDD metadata for the plugins:

kind: LifecyclePluginHook
version: v1
definition:
name: allow list
url: http://localhost:8787
pre: parse
config:
  request:
    headers:
      additional:
      hasura-m-auth:
        value: "your-strong-m-auth-key"
    session: {}
    rawRequest:
      query: {}
      variables: {}

The pre-parse plugin hook's request can be customized using the LifecyclePluginHook metadata object. Currently we support the following customizations:

• adding/removing session information
• adding new headers
• forwarding specific headers
• adding/removing graphql query and variables

Fixed

• Disallow model filter boolean expressions having relationship comparisons in their nested object filters.

Changed

v2024.08.07

Added

• A new CLI flag (--export-traces-stdout) and env var (EXPORT_TRACES_STDOUT) is introduced to enable logging of traces to STDOUT. By default, logging is disabled.

Remote Relationships Predicates

We have significantly enhanced our permission capabilities to support remote relationships in filter predicates. It is important to note that the relationship source and target models should be from the same subgraph.

Example: API traces are stored in a separate database. Users should only be able to view traces of their own API requests.

kind: ModelPermissions
version: v1
definition:
  modelName: traces
  permissions:
    - role: user
      select:
        filter:
          relationship:
            name: User
            predicate:
              fieldComparison:
                field: user_id
                operator: _eq
                value:
                  sessionVariable: x-hasura-user-id

In the above configuration, a permission filter is defined on the traces model. The filter predicate employs the User remote relationship, ensuring the user_id field is equal to the x-hasura-user-id session variable.

• New NoAuth mode in auth config can be used to provide a static role and session variables to use whilst running the engine, to make getting started easier.

Fixed

• Fixes a bug where queries with nested relationship selection and filter predicates fail due to an issue with NDC relationship collection

• Reduce error for using nested arrays in boolean expressions to a warning to maintain backwards compatibility

• Fix use of object types as comparison operator arguments by correctly utilising user-provided OpenDD types.

• Fixes a bug where argument presets set in the DataConnectorLink were sent to every connector function/procedure regardless of whether the function/procedure actually declared that argument

• Fixes a bug where argument presets set in the DataConnectorLink were not sent to connector collections that backed Models

• Fixes a bug where the type of the argument name in the DataConnectorLink's argument presets was incorrect in the Open DD schema. It was ArgumentName but should have been DataConnectorArgumentName

• Fixes a bug where the check to ensure that argument presets in the DataConnectorLink does not overlap with arguments defined on Models/Commands was comparing against the Model/Command argument name not the data connector argument name

Changed

• Introduced AuthConfig v2. This new version removes role emulation in engine (allowRoleEmulationBy) field.

• Raise a warning when an invalid data connector capabilities version is used in in a DataConnectorLink and prevent the usage of incompatible data connector capabilities versions

• Models and commands that do not define all the necessary arguments to satisfy the underlying data connector collection/function/procedure now cause warnings to be raised. The warnings will be turned into errors in the future.

v2024.07.25

Fixed

• Ensured traceresponse header is returned

v2024.07.24

Added

• The metadata resolve step now emits warnings to let users know about soon-to-be deprecated features and suggest fixes.

Fixed

• Fixes a bug where boolean expressions passed as arguments would not be translated into NDC Expression types before being sent to the data connector.

• Fixes a bug where relationships within nested columns would throw an internal error. While generating NDC relationship definitions, engine would ignore columns with nested selection.

• Renamed the ArgumentPreset for data connectors to DataConnectorArgumentPreset to avoid ambiguity in generated JSONSchema.

Changed

• Fixed a bug where command targeted relationships were not using the Open DD argument name instead of the data connector's argument name when querying the data connector

v2024.07.18

Added

Remote Relationships in Query Filter

We have enhanced the GraphQL query capabilities to support array and object relationships targeting models backed by different data connectors. This allows you to specify remote relationships directly within the where expression of your queries.

Example: Retrieve a list of customers who have been impacted by cancelled orders during the current sale event. This data should be filtered based on order logs stored in a separate data source.

query CustomersWithFailedOrders {
  Customers(
    where: {
      OrderLogs: {
        _and: [
          { timestamp: { _gt: "2024-10-10" } }
          { status: { _eq: "cancelled" } }
        ]
      }
    }
  ) {
    CustomerId
    EmailId
    OrderLogs {
      OrderId
    }
  }
}

By incorporating remote relationships into the where expression, you can seamlessly query and filter data that spans across multiple data sources, making your GraphQL queries more versatile and powerful.

Fixed

• Build-time check to ensure boolean expressions cannot be built over nested array fields until these are supported.

• Fixed a bug where command targeted relationships were not using the OpenDD argument name instead of the data connector's argument name when querying the data connector.

v2024.07.10

Fixed

• Fixes a bug with variable nullability coercion. Specifically, providing a non-null variable for a nullable field should work, as all non-nullable variables can be used as nullable variables via "coercion".

• Fixes a bug where data connectors without the foreach capability were not allowed to create local relationships

v2024.07.04

Added

• Query Usage Analytics - usage analytics JSON data is attached to execute span using internal.query_usage_analytics attribute

• Added a flag, --partial-supergraph, which instructs the metadata resolver to prune relationships to unknown subgraphs rather than failing to resolve

Boolean Expression Types

A new metadata kind BooleanExpressionType can now be defined. These can be used in place of ObjectBooleanExpressionType and DataConnectorScalarRepresentation, and allow more granular control of comparison operators and how they are used.

kind: BooleanExpressionType
version: v1
definition:
  name: album_bool_exp
  operand:
    object:
      type: Album
      comparableFields:
        - fieldName: AlbumId
          booleanExpressionType: pg_int_comparison_exp
        - fieldName: ArtistId
          booleanExpressionType: pg_int_comparison_exp_with_is_null
        - field: Address
          booleanExpressionType: address_bool_exp
      comparableRelationships:
        - relationshipName: Artist
          booleanExpressionType: artist_bool_exp
  logicalOperators:
    enable: true
  isNull:
    enable: true
  graphql:
    typeName: app_album_bool_exp

kind: BooleanExpressionType
version: v1
definition:
  name: pg_int_comparison_exp
  operand:
    scalar:
      type: Int
      comparisonOperators:
        - name: equals
          argumentType: String!
        - name: _in
          argumentType: [String!]!
      dataConnectorOperatorMapping:
        - dataConnectorName: postgres_db
          dataConnectorScalarType: String
          operatorMapping:
            equals: _eq
  logicalOperators:
    enable: true
  isNull:
    enable: true
  graphql:
    typeName: app_postgres_int_bool_exp

• Add flag to (--expose-internal-errors) toggle whether to expose internal errors. (#759)

Aggregates of Array Relationships

Aggregates of array relationships can now be defined by specifying an aggregate in the Relationship's target. Note that this is only supported when the target of the relationship is a Model. You must also specify the aggregateFieldName under the graphql section.

kind: Relationship
version: v1
definition:
  name: invoices
  sourceType: Customer
  target:
    model:
      name: Invoice
      relationshipType: Array
      aggregate: # New!
        aggregateExpression: Invoice_aggregate_exp
        description: Aggregate of the customer's invoices
  mapping:
    - source:
        fieldPath:
          - fieldName: customerId
      target:
        modelField:
          - fieldName: customerId
  graphql: # New!
    aggregateFieldName: invoicesAggregate

• One can now configure the engine to set response headers for GraphQL requests, if NDC function/procedures returns headers in its result

Field arguments

Add field arguments to the OpenDD ObjectType:

kind: ObjectType
version: v1
definition:
  name: institution
  fields:
    - name: id
      type: Int!
    - name: name
      type: String!
      arguments:
        - name: hash
          argumentType: String
        - name: limit
          argumentType: Int
        - name: offset
          argumentType: Int
  graphql:
    typeName: Institution
  dataConnectorTypeMapping:
    - dataConnectorName: custom
      dataConnectorObjectType: institution
      fieldMapping:
        id:
          column:
            name: id
        name:
          column:
            name: name
            argumentMapping:
              hash: hash
              offset: offset
              limit: limit

Changed

Fixed

• Engine now respects relation_comparisons capability, when generating GraphQL schema for relationship fields in model filter
• The OpenDD schema for DataConnectorLink now references the correct version (v0.1.4) of the NDC schema when using the NDC CapabilitiesResponse and SchemaResponse types

v2024.06.13

Initial release.