see releases
https://github.com/sangria-graphql/sangria/releases/tag/v3.0.0
https://github.com/sangria-graphql/sangria/releases/tag/v2.1.3
https://github.com/sangria-graphql/sangria/releases/tag/v2.1.2
https://github.com/sangria-graphql/sangria/releases/tag/v2.1.1
https://github.com/sangria-graphql/sangria/releases/tag/v2.1.0
https://github.com/sangria-graphql/sangria/releases/tag/v2.0.1
- can disable expensive suggestions (#517)
- Faster suggestion list (#521)
https://github.com/sangria-graphql/sangria/releases/tag/v2.0.0
- scala 2.13 support #1 on temporary fork
- Removed
InterfaceMustHaveImplementationValidationRule
validation rule (spec change) (#379). Since its introduction in previous release, it caused some issues (to sangria users as well users of other implementations). So it was removed from the GraphQL spec and sangria for now. - Exposed more contextual information to a fetcher, including fetcher cache (#377). Fetcher now can be created with a set of new helper methods
*WithContext
which provideFetcherContext
as an argument to fetch functions. AstSchemaMaterializer
now re-creates existing field argument types (thus able to use newly created input types)- Added visitor helpers in
AstNode
(they just delegate all functionality toAstVisitor
) - More minor improvements for better compatibility with GraalVM
native-image
. - Continued work on GraphQL CATs (Compatibility Acceptance Tests). Most recent changes and validation scenarios were added. Some
Violation
s now implementSpecViolation
which provides CATSs-compliant errorcode
and arguments.
-
Added support for
extend schema
in SDL (spec change) (#362). It adds following syntax in the SDL:extend schema @extraDirective { mutation: MutationType }
-
Ambiguity with null variable values and default values (spec change) (#359).
-
Add optional 'extensions' entry to errors (spec change) (#358).
CAUTION: breaking change. All additional error object fields that were provided via
HandledException
are now added to theextensions
fieldBefore v1.4.1:
{ "data": { "books": null }, "errors": [{ "message": "Can't access the database :'(", "path": ["books"], "locations": [{"line": 3, "column": 11}], "errorCode": "DATABASE_DOWN", "mitigationStrategy": "Panic!" }] }
After v1.4.1:
{ "data": { "books": null }, "errors": [{ "message": "Can't access the database :'(", "path": ["books"], "locations": [{"line": 3, "column": 11}], "extensions": { "errorCode": "DATABASE_DOWN", "mitigationStrategy": "Panic!" } }] }
For backward-compatibility,
HandledException
now provides 2 additional fields:addFieldsInExtensions
(by defaulttrue
) andaddFieldsInError
(by defaultfalse
). You can also set both flags totrue
in order to provide a migration path for the client applications. -
Ensure interface has at least 1 concrete type (spec change) (#360). It is potentially a minor breaking change. If you have non-implemented interfaces in your schema, you can temporary remove
InterfaceMustHaveImplementationValidationRule
schema validation rule. -
Added a small
Cache
abstraction and replacedTrieMap
-based cache implementation withConcurrentHashMap
. This change introduces potential minor performance improvements and compatibility with GraalVMnative-image
. -
Added
toAst
helper for different schema elements, like fields, enum values and types (#367) -
Added macro setting to transform enum values' names with a macro setting (#350). Big thanks to @fehu for this contribution! The
UppercaseValues
macro setting is now deprecated in favour of more flexibleTransformValueNames
. -
Fixed typo in
BeforeFieldResult
field name (#363). Big thanks to @Codier for this contribution! -
Adjusted
KnownTypeNames
validation to not validate SDL definitions because they are validated in the schema materializer (#354). -
Added option to disable parsing of AST locations and comments (#357)
-
Don't allow
ObjectType
s derived from case classes to overwrite each other (#345) Big thanks to @simonsaffer for this contribution! -
Improved query parsing performance by optimizing the AST location tracking (#351). Big thanks to @guymers for this contribution!
-
Fixed new line tracking when parsing block-strings
-
Ast schema builder now considers
additionalTypes
when it validates the type extensions -
Updated
sangria-marshalling-api
to version 1.0.3 (should be backwards compatible with v1.0.0).
The v1.4.0 a lot of improvements and minor changes, in particular around SDL parsing and materialization. Although some of these changes a minor breaking changes, they can be divided in following 2 categories:
- Simple routine refactorings with updated method signatures. For example most of the method signatures in
AstSchemaBuilder
are updated and now include more information (type extensions in particular). In all of these cases issues will manifest themselves as compilation errors and can be fixed by simple signature updates. - GraphQL query and SDL syntax updates. Based on the feedback from #308 and other sources, extra attention was put in ensuring that syntax is either backwards compatible or an option is available to enable support for a legacy GraphQL syntax. These options can be enabled though new
ParserConfig
which can be provided toQueryParser
.
If you are facing unexpected issues with migration to the new Sangria version, please let us know by creating a new GH issue.
Changes:
-
New "implements" syntax (old syntax can be enabled via
ParserConfig.legacyImplementsInterface
) (#337) (spec change). Example:# old syntax type User implements Node, Profile { id: ID! } # new syntax type User implements Node & Profile { id: ID! }
-
All of the SDL types now support type extensions (#337) (spec change). Previously only object types supported this feature. Some examples:
extend type Foo implements ExtraInterface1 & ExtraInterface2 { extraField: Int } extend interface Bar @extraDirective { extraField: Int } extend input Baz { extraField: Int = 123 } extend enum Color { "the best color" MAGENTA } extend union Test @extraDirective = More | Types extend scalar Date @extraDirective
-
All SDL type definitions are now allowed to have empty field/value/member lists (#337) (spec change). Examples of now valid syntax:
type User union Pet enum Color
All of the appropriate checks are now implemented as a
SchemaValidationRule
(instead of being part of the syntax).Moreover, old syntax for this is now deprecated and can be enabled with
ParserConfig.legacyEmptyFields
. Example of the old syntax:# don't do it! it is now deprecated type User {}
-
Add experimental support for parsing variable definitions in fragments (#313). Primary goal of it is to enable the experimentation. The support for this syntax needs to be explicitly enabled via
ParserConfig
. Here is how new syntax look like:query q () { ...a } fragment a on t ($size: Int = 0) { f(size: $size) }
-
Improved AST location handling and support for multi-source AST
Document
s (#343). These improvements might be quite helpful for recently discussed "import" functionality where the schema SDL is defined in multiple files. Improvements include:- Different schema elements retain the information about SDL
AstNode
s that were used to create them (in addition to the directives). This might be quite helpful in multiple scenarios. It was already used to greatly improve the quality of the error messages. AstLocation
now replaces thePosition
and holds additional field:sourceId
.- Introduced
AggregateSourceMapper
and improvedDocument
merge. It is now possible to show proper error messages and source locations even for AST that was parsed from different sources (files, URLs, etc.).
- Different schema elements retain the information about SDL
-
A lot of improvement to the schema validation (#312, #315). All of the validations are now implemented as
SchemaValidationRule
(no in-place runtime checks anymore). This can be quite helpful if you, for instance, need to customize the validation rules or disable the validations altogether. -
Validate literals in a single rule with finer precision (#314). This generalizes the "arguments of correct type" and "default values of correct type" to a single rule "values of correct type" which has been re-written to rely on a traversal rather than the utility function
isValidLiteralValue
.isValidLiteralValue
is now deprecated. -
AstSchemaMaterializer
now fully supports all of the new type extensions (#309) . -
SchemaComparator
is improved and now includes information about "dangerous" changes (#335). -
Simplify Unknown Args Validation (#316).
-
Added new validation rule
SingleFieldSubscriptions
(#254) (spec change). -
Resolve type info for fragment spreads (#278). Big thanks to @jonas for this contribution!
-
Allow to rename, describe and set default arguments using
DeriveObjectSetting
s (#339). Big thanks to @fehu for this contribution! -
Replaced Exception with Throwable to match all possible results of Future.failed (#329, #327). Big thanks to @lgmyrek for this contribution!
-
Initialize symbols before checking their annotations (#317). This will improve
derive*
macros in same edge-cases. Big thanks to @dragos for this contribution! -
Added support for rendering SDL with legacy comment-based descriptions (#334). You can use it with
SchemaFilter.default.withLegacyCommentDescriptions
. -
Improved
ResolverBasedAstSchemaBuilder
and introducedInputTypeResolver
/OutputTypeResolver
. -
Updated
sangria-marshalling-api
to version 1.0.1 (should be backwards compatible with v1.0.0) -
Other minor improvements coming from the reference implementation (#336, #311).
-
Removed previously deprecated methods:
SchemaRenderer.renderIntrospectionSchema
AstVisitor.visitAst
-
Added support for string-based descriptions in SDL and implement most recent SDL-related changes (#303, #304, #305) (spec change). Breaking change for all SDL users. The type, field and enum value descriptions are now handled differently in the SDL.
Old format (comment-based):
# This is a description # of the `Foo` type. type Foo implements Bar { # another description one: Type }
New format (string-based):
""" This is a description of the `Foo` type. """ type Foo implements Bar { "another description" one: Type }
Migration path: by default (and after the update) only new description format is used. The old (comment-based) format is still will be supported for several version. In order to enable description handling in the old format, please override
DefaultAstSchemaBuilder.useLegacyCommentDescriptions
or useAstSchemaBuilder.defaultWithLegacyCommentDescriptions
. -
Added block string support (#260, #301) (spec change). Here is an example of the new multi-line string syntax:
mutation { sendEmail(message: """ Hello, World! Yours, GraphQL. """) }
-
Added
relOnly
andrelOnlyCaching
methods toFetcher
. Thanks to @kuppuswamy for this contribution! -
Fixed a validation of nested InputObjectType with same property names (#292).
-
Improved performance of
Document.hashCode
which had a big influence on performance of the AST-based schema builder.
- Fix regression where variables aren't passed to query reducers (#291). Thanks to @msolomon for a quick fix!
-
High-level API for SDL-based schema materialization (#288). It provides much more simple and robust API for building an executable schema based on SDL definitions. For more info see "High-level SDL-based Schema Builder" section of the documentation.
As a part of this feature, new functionality was introduced in
AstSchemaBuilder
which provides even more flexibility for schema materialization.Minor breaking change: most of the methods in
AstSchemaBuilder
got new argumentorigin: MatOrigin
. In order to migrate, you need to adjust the signatures of affected methods and addorigin: MatOrigin
argument. -
Prepared queries without known variables (#281, #277). Huge thanks to @msolomon for making this contribution! This feature adds
QueryReducerExecutor.reduceQueryWithoutVariables
. In its signature it is similar toExecutor.prepare
, but it does not requirevariables
and designed to validate and execute query reducers for queries that are being analyzed ahead of time (e.g. in the context of persistent queries). -
Ability to provide a partial error for deferred values (#290).
DeferredValue
andDeferredFutureValue
now has a methodmapWithErrors
. This works similar to thePartialValue
, so you can still have a successful result and at the same time indicate errors that happened during deferred value resolution. -
Ability to provide additional values from Middleware (#289).
beforeField
now returnsBeforeFieldResult
(instead of just a tuple) which allows you to add aMiddlewareAttachment
. This attachment then can be used in resolve function viaContext.attachment
/Context.attachments
.
-
Experimental Batch Executor (#273). For more info see the "Batch Executor" section of the documentation.
-
Allow
Violation
s andUserFacingError
s to be handled by custom exception handler (#252).Refactored exception handling mechanism:
- Now it is able to handle
Violation
s as well asUserFacingError
s. HandledException
is now able to capture multiple errors and additional AST node positions.- Since it is now a standalone class, it would be easier to expand on error handling in future.
Minor breaking change. It's just a small syntax change. Migration strategy:
// Before val exceptionHandler: Executor.ExceptionHandler = { case (m, ...) => ... }) // After val exceptionHandler = ExceptionHandler { case (m, ...) => ... }
ExceptionHandler
is now a standalone class that allows you to provide following handlers:onException
- all unexpected exceptions coming from theresolve
functions (behaves exactly like in earlier versions)onViolation
- handles violations (things like validation errors, argument/variable coercion, etc.)onUserFacingError
- handles standard sangria errors (errors like invalid operation name, max query depth, etc.)
For more info see the updated "Custom ExceptionHandler" section of the documentation.
- Now it is able to handle
-
Improved input document validation and deserialization (#272). For more info see "Input Document Validation" section of the documentation and updated "Query AST Marshalling" section. Improvements include:
- Added
InputDocument
which is used in validation, materialization, etc. - New macros
gqlInpDoc
/graphqlInputDoc
that produces anInputDocument
instead of justsangria.ast.Value
. - Added
RuleBasedQueryValidator.validateInputDocument
that validatesInputDocument
against the schema. InputDocument.to
provide a convenient way todeserialize
/materialize
an input document based on theFromInput
type-class.- Improved a lot of validation messages related to input value validations.
- Added
-
Add support for leading vertical bar in union types and directive definitions (#253) (spec change).
-
Fixed infinite loop on invalid queries in OverlappingFields (#266, #238).
-
Information about type extensions is now available in the field resolve function builder (#267). Minor breaking change. The signature of
DefaultAstSchemaBuilder.buildField
andDefaultAstSchemaBuilder.resolveField
has changed. You need to addextensions: Vector[ast.TypeExtensionDefinition]
as a second argument. -
Fixed directive definition rendering in query renderer (#274). Thanks to @alexeygolev for this contribution!
-
Built-in scalars will now only be added to the schema if they are used (#271, #270). Thanks to @jlawrienyt for this contribution!
-
Improve error message when an appropriate implementation of an abstract type cannot be found (#259).
- Added new middleware traits
MiddlewareFromScalar
andMiddlewareToScalar
. They provide a way to transform all scalar values from middleware (#249, #248). This have some advantages since middleware can be disable, chained together and has access to context information. Huge thanks to @BjRo and @Axxiss for helping with the feature design and implementation! - Added new middleware trait
MiddlewareExtension
(#256). It provides an easy way to add extensions from middleware. - Improved error message for
All fields within a Type should have unique names!
(#247). It now includes type and field information. - Fixed helper methods for operation lookup in
ast.Document
.
- Fixed
MeasureQueryDepth
reducer not keeping the largest depth found (#245, #246). Big thanks @Eluinhost for this contribution! - Easier way to create
Args
(#243). Big thanks @vishalkuo for this contribution! - More options to render a schema (#241). This improvement is especially useful for apps that use relay modern.
- Provide convenient functions for IDL-based schema materialization and validation (#240). For more info see the "Query Validation" section of documentation. Improvements include:
- Introduced Schema.buildStubFromAst that builds a schema with a stub Query type
- Introduced Schema.buildDefinitions that builds a list of definitions based on IDL AST (without a schema)
- Introduced Document.emptyStub as a most basic, but valid document stub
- Introduced alias query1 + query2 for document merge
- Add
Fetcher.deferSeqOptExplicit
or similar to explicitly getSeq[Option[T]]
in the result (#230) - Fixed scalar aliases when they are used with variables or schema is extended (#237)
- Preserve IDL directives at schema materialization time and use them in schema comparator (#236). This also adds
Vector[ast.Directive]
to all relevant schema definitions which may be extremely useful for future features and schema analysis - Improve syntax error reporting for
graphql
macro (#235) - Improve
Int
,BigInt
andLong
scalar value handling (#234) - Propagate updated value through middleware's
afterField
(#233). For more info see the "Middleware" section of documentation. - Forbid 'true', 'false' and 'null' as names for Enum value (#239)
- Added scalar type alias support (#225, #210). For more info see the "Scalar Type Alias" section of documentation.
- Greatly improved
AstVisitor
(#214). It now includes helper methods to traverse and transform an AST with type info and state. Sangria now integrates with macro-visit which was specifically written to traverse and transform ASTs which are similar to sangria's. For more info see the "AstVisitor" section of documentation. - Added
DocumentAnalyzer
andSchemaBasedDocumentAnalyzer
that contain a lot of helper methods to analyze query. This includes newly introduceddeprecatedUsages
andintrospectionUsages
(#211, #207, #212). For more info see brand new "Query And Schema Analysis" section of documentation. - Added
QueryReducer.hasIntrospection
andQueryReducer.rejectIntrospection
that rejects queries which contain introspection fields (#211). This may be useful for production environments where introspection can potentially be abused. - Added
Context.isIntrospection
andsangria.introspection.isIntrospection
helper methods to easily distinguish between introspection types and fields in middleware and query reducers. - Added
QueryReducer.measureDepth
andQueryReducer.rejectMaxDepth
rejects queries that are deeper than provided threshold (In contrast toExecutor.execute(maxQueryDepth = ...)
, query reducer does it at query analysis time before actual execution). - Added
derive*
macro settings (TransformFieldNames
andTransformInputFieldNames
) to transform field names. (#215) Big thanks to @ostronom for this contribution. - Added ability to represent complex relations in Fetch API (#220).
ExecutionScheme.Extended
now returns updated user context for mutations (#209).- Improved handling of tailing comments when rendering query AST (#219).
- Added aliases for
graphql
/graphqlInput
macros:gql
/gqlInp
- Using
Vector
instead ofList
for all AST nodes now. This is a minor breaking change.
- Added
Action.sequence
combinator to compose a list ofLeafAction
s in a singleLeadAction
that can be returned from aresolve
function (#206) - Support of
Option[Id]
inFetcher.deferOpt
(#205) - Implicit conversion from
Future[Deferred[A]]
toAction
does not work (#201) - Disallow creation of object types with empty list of fields (#200)
- SimpleFetcherCache does not cache Relation (#194)
- Fetching Relation Typing (#193)
- Helper method
sangria.schema.action
is replaced withAction.apply
andLeafAction.apply
- Uphold spec for non-validation names not beginning with
__
(spec-change) (#189) - Field cache does not consider output object polymorphism (#190)
- Added
QueryReducer.measureDepth
andQueryReducer.rejectMaxDepth
query reducers (#191).maxQueryDepth
argument was available for a long time onExecutor
, but in contrast to new query reducers, it measures depth along side of the execution. With query reducers it happens before query execution, thus allow to reject the query execution entirely. - FetcherDeferredOpt does not extends DeferredOpt (#188)
- Invalid operation name now be considered a client-side error and now implements
QueryAnalysisError
(#186, #187). Big thanks to @mattfenwick for working on this one!
Sangria v1.0.0-RC4 is fully compliant with "October 2016" version of the GraphQL specification.
- In presence of explicit
null
value, the default will not be used (#185) (spec change). Breaking change! Arguments with default values are no longer treated as non-optional arguments. ThoughArgs
still preserves existing semantics (default is still applied, even in presence of explicitnull
). The same is true for input objects and optional input object fields with default values. - Added
Args.withArgs
andArgs.argDefinedInQuery
convenience methods. - Validation rule for unique directives per location (#179).
- Enforces input coercion rules (#177).
MiddlewareQueryContext
andExecutionResult
now contain information about validation and query reducers' execution time (validationTiming
,queryReducerTiming
)Middleware.fieldError
was not called in all possible exceptional situations (#184).- New
QueryParser.parseInputWithVariables
provides a way to parse input values with variables. - Various bugfixes in error handing of deferred values and null/undefined value handling.
- Cross compile to scala 2.12 and scala 2.11 (#182, #183)
- Schema comparator (#169, #165). It helps to compare different schemas or different versions of the same schema. It also provides an information whether particular change is a breaking change. This is a great example of GraphQL type system potential. For more info see the "Schema Comparator" section of documentation.
- Improve handling of
NaN
andinfinity
values (#167, #168)
- Capture original exceptions only if necessary (based on the
ExecutionScheme
) - Fixed issue with duplicate errors appearing during sequential query execution (mutations)
Towards 1.0!
- Stream-based subscriptions (#98, #166). For more info see the "Stream-based Subscriptions" section of documentation.
- High-level Fetch API for deferred value resolution (#164). For more info see the "High-level Fetch API" section of documentation.
- Huge improvements in deferred value resolution (#161). Here is just an example of how batching algorithm is improved in comparison to previous version. For more info see the "Deferred Value Resolution" section of documentation. It got a lot of new content.
- Introduced
ExecutionScheme
. It allows to change the result type of an exaction. So now you can get some meta-information about a query execution itself, and not only theFuture
of marshaled result. For more info see the "Alternative Execution Scheme" section of documentation. - Minor breaking changes:
-
DeferredResolver
andDeferred
are moved tosangria.execution.deferred
package. -
DeferredResolver.resolve
method signature is changes a bit (2 new arguments were added). Here is the new signature:def resolve(deferred: Vector[Deferred[Any]], ctx: Ctx, queryState: Any)(implicit ec: ExecutionContext): Vector[Future[Any]]
-
- Description formatting/parsing is updated based in the changes in the reference implementation (#155). More places of the query now preserve the comments. For instance all trailing comments within a section set and at the end of the document are preserved and rendered.
- Ensure that the result of deferred value resolution has the same size as the deferred list (#154).
- During macro-based derivation, default value should never be a
null
for an optional arguments (#153). - Executor now properly handles undefined values (like
null
andNone
) even if GraphQL type is not null (#152)
- Transitive types are now collected for all types provided via
additionalTypes
to a schema definition (#149). ObjectType.withInstanceCheck
provides an easier way to customize an instance check onObjectType
(#148).Enumeration
derivation macro should now only collects instances ofEnumeration#Value
(#151).- Ensure that all unreferenced types are collected during the schema extension.
sangria.ast.Type.namedType
helper method.
-
Provide
extendSchema
utility function (#113). This feature allows you to extend existing schema with additional types and existing types with additional fields. It may be very useful for client-side tools and for server implementations in cases where parts of a schema are dynamically generated or coming from external sources (like database).Here is a small example of how you can use it:
val schema: Schema[Ctx, Val] = Schema(...) val schemaExtensions = graphql""" extend type Human { pet: Animal @loadPetInfo } interface Animal { name: String! } type Dog implements Animal { name: String! nickname: String } type Cat implements Animal { name: String! age: Int } """ val myCustomBuilder = new DefaultAstSchemaBuilder[Ctx] {...} val extendedSchema = schema.extend(schemaExtensions, myCustomBuilder)
Just like with AST-based schema materialization, you can provide a custom schema builder which allows you to control most of the aspects of generated parts of the schema.
-
Handling of more than one
ProjectionName
for one field (#146). -
Updated context propagated only to siblings (#145).
- Initial CATs (Compatibility Acceptance Tests) support (#142). The test suite itself is still work-in-progress, but sangria includes an integration which executes all currently available test cases.
- IDL (schema definition) syntax parsing and rendering (#137, #62)
- AST-based schema materializer (#139, #115). This feature may be very useful for different tools that need to create an executable schema based on IDL definitions. Almost any aspect of generated in-memory schema representation can be customized via custom implementation of
AstSchemaBuilder
. This feature is already used in sangria itself for CATs (Compatibility Acceptance Tests) integration. At the moment default implementation ofAstSchemaBuilder
treats comments that start with##
as a field/object/argument description. - Partial resolve
Action
s (#140). This change introduces 2 newAction
s that can be returned back from aresolve
function:PartialValue
andPartialFutureValue
. This allows you to return a list of errors in addition to a successfully resolved value (which may contain only partial result due to the errors). - Preserve comments during the AST parsing (#105). Most of the AST classes got comment
Option[Comment]
field. It can be very useful for query formatting becauseQueryRenderer
also got support for comments and able to render them. - Include execution path in error objects (#143). This may be helpful for client side tools that would like to analyze error messages and programmatically use them in some way. This is a minor braking change since
field
property on error is removed in favor of newpath
property which is a list. - Introspection-based schema materializer now also uses more advanced
IntrospectionSchemaBuilder
(similar to the AST-based one) instead ofMaterializationLogic
, which is now removed. This introduces a minor breaking change, but in a long runIntrospectionSchemaBuilder
will provide much more flexibility. - Add comment/directive support in the introspection-based schema renderer (#136).
- Validation: improving overlapping fields quality (#133)
- Deprecated directive (#132)
- New directive locations (#131)
- Default values should be in GraphQL format (introspection) (#141)
- Added support for case objects defined in companion object (#135). Big thanks to @joprice for contributing this improvement!
SchemaRenderer
now has improved default value rendering- Execution
path
now got it's own classExecutionPath
(which is now used instead of simpleVector[String]
). This introduces a minor breaking change.
-
Marshaling for Amazon Ion data format is introduced. Amazon Ion is a richly-typed, self-describing, hierarchical data serialization format offering interchangeable binary and text representations.
You need following dependency to use it:
"org.sangria-graphql" %% "sangria-ion" % "0.1.0"
In order to use Ion marshalling, you need an implicit instance of
IonSystem
in scope as well:import sangria.marshalling.ion._ implicit val ionSystem = IonSystemBuilder.standard().build() val result: Future[IonValue] = Executor.execute(schema, query)
-
Marshalling API is updated to v0.2.1. It introduces a minor breaking change. This change introduces performance improvements to scalar value marshalling and gives much more flexibility in terms of the type of marshaled values.
ResultMarshaller
now able to communicate it's natively supported capabilities to aScalarType
viaMarshallerCapability
. A set of standard marshaller capabilities were introduced:DateSupport
- Marshaller supportsjava.util.Date
natively.CalendarSupport
- Marshaller supportsjava.util.Calendar
natively.BlobSupport
- Marshaller supports large binary objects in form ofArray[Byte]
natively.
This still requires you to create a custom scalar types (for dates, blobs, etc.), but it gives you an ability to generically use native features of underlying data format.
ScalarType
now also able to communicate back to marshaller viaScalarValueInfo
. This can be used, for instance, to represent anArray[Byte]
as aclob
type instead ofblob
in formats that support both of them (like Amazon Ion). -
Include possible field, argument, type names when validation fails (#126).
-
Deepen introspection query from 3 levels to 7 (#128).
-
Improve validation error message when field names conflict (#130).
-
Interface hierarchies are not correctly rendered with
SchemaRenderer
(#125). -
Updated parboiled to v2.1.3
This release is fully compatible with "April 2016" version of the GraphQL specification.
- Return type overlap validation (#124) (spec change).
deriveContextObjectType
/deriveObjectType
do not work with Option arguments in some cases (#123)
A minor maintenance release to keep up with the spec changes.
- Field order in the result now reflects field order in the query (according to the spec) for all marshalling libraries that support field ordering (#99) (spec change).
- Directive
locations
field replacesonOperation
,onFragment
andonField
(#119) (spec change). - Low-level marshalling API is improved: it's now possible to use efficient map builders (which also able to preserver an order of the fields). This improves serialization performance and minimizes memory footprint. All marshalling libraries already take advantage of this API.
SchemaRenderer
prints duplicated fields for a type that implements an interface (#122)
-
Macro-Based GraphQL Type Derivation (#120). See "Macro-Based GraphQL Type Derivation" section of the documentation for more info.
-
Prepared Queries (#118). See "Prepared Queries" section of the documentation for more info.
-
Executor.execute
now returnsFuture
with failure if error happened before query execution (#109). It can be extremely helpful when you need to take some action or produce different result in case of error. Typical example is returning different HTTP status code.CAUTION: breaking change and action needed! Since things like validation errors and errors in query reducers are now explicitly returned as a
Future
failure and not as a successful result, you need to take some action to handle them. In order to migrate, all you need to do is to add followingrecover
:Executor.execute(schema, query).recover { case error: ErrorWithResolver => error.resolveError }
recover
function will make sure that all of the errors, that were previously handled internally inExecutor
, are now properly handled. Code above will produce exactly the same result as before.resolveError
produces a valid GraphQL response JSON and will use custom exception handler, if you have provided one.This new approach to error handling gives you much more flexibility. For example in most cases it makes a lot of sense to return 400 HTTP status code if query validation failed. It was not really possible to do this before. Now you able to do something like this (using playframework in this particular example):
executor.execute(query, ...) .map(Ok(_)) .recover { case error: QueryAnalysisError => BadRequest(error.resolveError) case error: ErrorWithResolver => InternalServerError(error.resolveError) }
This code will produce status code 400 in case of any error caused by client (query validation, invalid operation name, etc.).
Errors that happened in a query reducer would be wrapped in
QueryReducingError
. Here is an example of returning custom status code in case of error in the query reducer:val authReducer = QueryReducer.collectTags[MyContext, String] { case Permission(name) => name } { (permissionNames, ctx) => if (ctx.isUserAuthorized(permissionNames)) ctx else throw AuthorizationException("User is not authorized!") } Executor.execute(schema, queryAst, userContext = new MyContext, queryReducers = authReducer :: Nil) .map(Ok(_)) .recover { case QueryReducingError(error: AuthorizationException) => Unauthorized(error.getMessage) case error: QueryAnalysisError => BadRequest(error.resolveError) case error: ErrorWithResolver => InternalServerError(error.resolveError) }
HTTP status code would be 401 for unauthorized users.
If you have issues with the migration, please raise an issue so that we can find a good solution together.
-
Minor breaking change.
userContext
androot
arguments ofExecutor
are moved inExecutor.execute
method. -
Detect name collisions with incompatible types during schema definition (#117)
-
Introduced a type alias
Executor.ExceptionHandler
for exception handler partial function
-
Added introspection-based schema materializer (#21). This feature has a lot of potential for clint-side tools, testing, mocking, creating facade GraphQL servers, etc.
Here is simple example of how you can use this feature (Using circe in this particular example):
import io.circe._ import sangria.marshalling.circe._ val introspectionResults: Json = ??? // coming from other server or file val clientSchema: Schema[Unit, Unit] = Schema.buildFromIntrospection(introspectionResults)
It takes a results of full introspection query (loaded from the server, file, etc.) and recreates the schema definition with stubs for resolve methods. You can customize a lot of aspects of materialization by providing custom
MaterializationLogic
implementation (you can also extendDefaultMaterializationLogic
class). This means that you can, for instance, plug in some generic field resolution logic (resolveField
method) or provide generic logic for custom scalars (coerceScalar*
methods). Without these customisations schema only would be able to execute introspection queries.By default, default values (for input object fields and arguments) would be ignored because it's just a string as far as introspection API is concerned. However you can enable default value support if you know the format of the default values (in many cases it would be JSON). There is even a helper function for this:
import spray.json._ import sangria.marshalling.sprayJson._ val clientSchema: Schema[Unit, Unit] = Schema.buildFromIntrospection(introspectionResults, MaterializationLogic.withDefaultValues[Unit, JsValue])
This will inform schema materializer that default values are serialized as JSON and that spray-json should be used to work with them (please note, that circe does not have a built-in JSON parsing support, so it can't be used out-of-the-box here. On the other hand, it's pretty easy to add support for particular circe parser by defining an implicit instance of
InputParser
type class). -
SchemaRenderer.renderSchema
is now able to renderSchema
objects and only introspection results (#114). This can be useful if you already have schema in memory and don't want to execute an introspection query against the schema in order to render it. -
Query validation rule: Unique variable names (#112)
-
Add suggested types to incorrect field message (#111)
-
Introspection result now has a parser which deserializes a JSON (or any other format) to a set of case classes. This may simplify client-side tools that work with introspection queries. Please use
sangria.introspection.IntrospectionParser.parse
to parse an introspection query results. -
Introduced
InputParser
type class in order provide optional support for default value parsing in schema materialization. -
Updated descriptions of a scalar values
-
Updated dependencies
-
Minor improvements
-
JSON library integration is extracted to separate libraries (#38). Evey integration library will have a separate and independent versioning and release cycle. Following new libraries were introduced:
-
sangria-marshalling-api now includes all of the interfaces that marshalling library needs to implement.
-
sangria-marshalling-testkit contains a set of generic test cases that can be used to test a concrete marshalling library integration.
-
sangria-spray-json contains an integration with spray-json library. From now on, please use following dependency if you would like to use spray-json support:
libraryDependencies += "org.sangria-graphql" %% "sangria-spray-json" % "0.1.0"
The package is changed for the sake of consistency. From now on please use following import:
import sangria.marshalling.sprayJson._
-
sangria-play-json contains an integration with play-json library. From now on, please use following dependency if you would like to use play-json support:
libraryDependencies += "org.sangria-graphql" %% "sangria-play-json" % "0.1.0"
The package is changed for the sake of consistency. From now on please use following import:
import sangria.marshalling.playJson._
-
sangria-json4s-native contains an integration with json4s-native library. From now on, please use following dependency if you would like to use json4s-native support:
libraryDependencies += "org.sangria-graphql" %% "sangria-json4s-native" % "0.1.0"
The package is changed for the sake of consistency. From now on please use following import:
import sangria.marshalling.json4s.native._
-
sangria-json4s-jackson contains an integration with json4s-jackson library. From now on, please use following dependency if you would like to use json4s-jackson support:
libraryDependencies += "org.sangria-graphql" %% "sangria-json4s-jackson" % "0.1.0"
The package is changed for the sake of consistency. From now on please use following import:
import sangria.marshalling.json4s.jackson._
-
sangria-circe contains an integration with circe library. From now on, please use following dependency if you would like to use circe support:
libraryDependencies += "org.sangria-graphql" %% "sangria-circe" % "0.1.0"
The package is changed for the sake of consistency. From now on please use following import:
import sangria.marshalling.circe._
-
-
Argonaut scala JSON library is now supported via sangria-argonaut (#59). Please use following dependency if you would like to use argonaut support:
libraryDependencies += "org.sangria-graphql" %% "sangria-argonaut" % "0.1.0"
And here is an import statement:
import sangria.marshalling.argonaut._
-
Added
operationType
andoperation
onast.Document
to easily identify the operation type (#110) -
Added a utility function to convert between different input representations (#108). This functionality is available though
sangria.marshalling.MarshallingUtil
.
-
A lot of performance improvements across the whole library
-
Added basic subscription support as defined in the spec (graphql/graphql-spec#109) and reference implementation (#89). At the moment subscriptions are pretty basic, so it's meant more for experiments rather than for use in real applications. It is very likely that this feature will experience breaking changes in the near future (spec change)
-
Much better handling of input objects (#37, #70). A new type-class is introduced:
FromInput
. It provides high-level and low-level way to deserialize arbitrary input objects, just likeToInput
.In order to use this feature, you need to provide a type parameter to the
InputObjectType
:case class Article(title: String, text: Option[String]) val ArticleType = InputObjectType[Article]("Article", List( InputField("title", StringType), InputField("text", OptionInputType(StringType)))) val arg = Argument("article", ArticleType)
This code will not compile unless you define an implicit instance of
FromInput
forArticle
case class:implicit val manual = new FromInput[Article] { val marshaller = CoercedScalaResultMarshaller.default def fromResult(node: marshaller.Node) = { val ad = node.asInstanceOf[Map[String, Any]] Article( title = ad("title").asInstanceOf[String], text = ad.get("text").flatMap(_.asInstanceOf[Option[String]]) } }
As you can see, you need to provide a
ResultMarshaller
for desired format and then use a marshaled value to create a domain object based on it. Many instances ofFromInput
are already provided out-of-the-box. For instanceFromInput[Map[String, Any]]
was added to support existing map-like data-structure format. All supported Json libraries also provideFromInput[JsValue]
so that you can use Json AST instead of working withMap[String, Any]
.Moreover, play-json and spray-json integration provide support for
Reads
andJsonFormat
. This means that your domain objects are automatically supported as long as you haveReads
orJsonFormat
defined for them. For instance this example should compile and work just fine without explicitFromInput
declaration:import sangria.integration.playJson._ import play.api.libs.json._ case class Article(title: String, text: Option[String]) implicit val articleFormat = Json.format[Article] val ArticleType = InputObjectType[Article]("Article", List( InputField("title", StringType), InputField("text", OptionInputType(StringType)))) val arg = Argument("article", ArticleType)
CAUTION: this is minor breaking change. Together with
null
value support, this feature changes the way input objects are deserialized into map-like structures (which still happens by default). Optional input fields will now produce input objects like:// for JSON input: {"op1": "foo", "opt2": null} Map("opt1" -> Some("foo"), "opt2" -> None) // for JSON input: {"op1": "foo"} Map("opt1" -> Some("foo"))
instead of (old format):
// for JSON input: {"op1": "foo", "opt2": null} Map("opt1" -> "foo") // for JSON input: {"op1": "foo"} Map("opt1" -> "foo")
As you can see, this allows you to distinguish between "undefined" json object fields and json object fields that are set to
null
. -
null
value support (as defined in the spec change: graphql/graphql-spec#83) (#55) (spec change) -
Extracted input value parsing and made it a first-class citizen (#103). So now you can parse and render any
ast.Value
independently from GraphQL query. There is even a newgraphqlInput
macros available:import sangria.renderer.QueryRenderer import sangria.macros._ import sangria.ast val parsed: ast.Value = graphqlInput""" { id: "1234345" version: 2 # changed 2 times deliveries: [ {id: 123, received: false, note: null, state: OPEN} ] } """ val rendered: String = QueryRenderer.render(parsed, QueryRenderer.PrettyInput) println(rendered)
It will print something like this:
{ id: "1234345" version: 2 deliveries: [{ id: 123 received: false note: null state: OPEN }] }
InputUnmarshaller
andResultMarshaller
are also now available for it, so you can useast.Value
as a variables or it can be a result of GraphQL query execution (instead of more traditional JSON). -
ToInput
,InputUnmarshaller
andResultMarshaller
are moved tosangria.marshalling
package. -
Improved error messages for input values (#86). Now they will contain the reason why particular value is invalid.
-
Implementations of interfaces can include additional field args (#90) (spec change)
-
Loosen overlapping field validation rules (#94) (spec change)
-
False positive validation error from fragment cycle when unknown fragment (#95)
-
Interfaces with covariant return types (#96)
-
A lot of minor changes and performance improvements in validation rules and query validator (#97) (spec change)
-
Add error handling in the
SchemaRenderer
(#100) -
Ambiguous implicit when using a
UnionType
bug (#101) -
A lot of internal refactorings (especially in variable and argument processing) to make everything above possible
QueryReducer
is introduced. It allows you to analyze a query and take an action before it's executed. It provides very similar functionality to complexity analysis (introduced in previous release), but in much more generic form. That's because complexity analysis is now rewritten as aQueryReducer
. In order to migrate, you need to replacemeasureComplexity
function withQueryReducer.measureComplexity
. Here is an example:Since rejection of complex queries is such a common use-case, there is now a helper function to create a reducer for it:val complReducer = QueryReducer.measureComplexity[MyCtx] { (c, ctx) => complexity = c ctx } Executor.execute(schema, query, userContext = new MyCtx, queryReducers = complReducer :: Nil)
val rejectComplexQuery = QueryReducer.rejectComplexQueries[MyCtx](14, (c, ctx) => new IllegalArgumentException(s"Too complex query: max allowed complexity is 14.0, but got $c"))
Middleware
got a type parameter forCtx
. This is a minor breaking change. If you don't use theuserContext
inside of theMiddleware
, then you can just parametrize it withAny
.- Complexity function on field should also be given the
Ctx
(#87)
- Query complexity calculation mechanism is implemented (#85). This mechanism makes a rough estimation of the query complexity before it is executed.
Every field in the query gets a default score
1.0
. The "complexity" of the query is the sum of all field scores. You can customize the field score withcomplexity
argument:If you would like to use this feature, you need to provideField("pets", OptionType(ListType(PetType)), arguments = Argument("limit", IntType) :: Nil, complexity = Some((args, childrenScore) => 25.0D + args.arg[Int]("limit") * childrenScore), resolve = ctx => ...)
measureComplexity
argument to theExecutor
. For example:The complexity of full introspection query (used by tools like GraphiQL) isval rejectComplexQueries = (c: Double) => if (c > 1000) throw new IllegalArgumentException(s"Too complex query: max allowed complexity is 1000.0, but got $c") else () val exceptionHandler: Executor.ExceptionHandler = { case (m, e: IllegalArgumentException) => HandledException(e.getMessage) } Executor.execute(schema, query, exceptionHandler = exceptionHandler, measureComplexity = Some(rejectComplexQueries))
102.0
. - json4s-jackson is now supported in addition to native (#84). This results in minor import change:
// before sangria.integration.json4s._ // after // either (same behaviour as before) sangria.integration.json4s.native._ //or sangria.integration.json4s.jackson._
- json4s is updated to version 3.3.0 (#84)
- Provide a helpful error messages if schema has a broken circular references (which cause
fields
to be null) (#83)
For the most part implemented spec changes. Now compatible with "October 2015" version of the GraphQL spec.
- Type condition optional on inline fragments. (#82) (spec change)
- Make operation name optional (#81) (spec change)
- Introspection descriptions for scalars and introspection (#80)
beforeField
now able to replace value and preventresolve
call (#79). This can be useful for things like caching. It contains minor breaking change - return type ofbeforeField
has changed. If you are implementing it, just returncontinue
if yourFieldVal
wasUnit
orcontinue(someFieldVal)
.Projection
andNoProjection
should be tags instead of resolve function wrappers (#78). Backwards-incompatible change: you need to replaceProjection
withProjectionName
tag andNoProjection
withProjectionExclude
tag. here is an example:// before Field("id", StringType, Some("The id of the droid."), resolve = Projection("_id", _.value.id)), // after Field("id", StringType, Some("The id of the droid."), tags = ProjectionName("_id") :: Nil, resolve = _.value.id)
This release contains quite a few backwards-incompatible changes, but fear not - all of them are renames and similar minor changes which should be easy to migrate. I collected all of them in the change list below. They were necessary in order to ensure consistent naming and improve the structure and flexibility of the library.
- #68 - Better handling of default input values. It's a part of ongoing effort to improve handling of input objects (#37). Default values should now have an instance
of
ToInput
type-class which is defined for all supported input types like scala map-like data structures, different json ASTs, etc. It even supports things likeWrites
from play-json orJsonFormat
from spray-json by default. This means that you can use your domain objects (likeUser
orApple
) as a default value for input fields or arguments as long as you haveWrites
orJsonFormat
defined for them. The mechanism is very extensible, of course: you just need to define implicitToInput[T]
for a class you want to use as a default value. This change makes it impossible to verify the default value type at compile time, since it can have any shape, like Json AST or maybe even some binary format. Don't worry though, at a schema creation time all default values would be validated according to the input type. - #77 - Middleware support. This addition has a huge potential: you can measure performance, collect metrics, enforce security, etc. on a field and query level. Moreover
it makes it much easier for people to share standard middleware in a libraries (e.g. sangria-security, sangria-graphite, sangria-influxdb, etc.). In order to ensure generic classification of
fields, every field now got a generic list or
FieldTag
s which allow to provide user-defined meta information about this field (just to highlight a few examples:Permission("ViewOrders")
,Authorized
,Measured
, etc.). You can find more info in docs and auth example - #76 - You can now provide
maxQueryDepth
toExecutor
. It will then enforce this constraint for all queries (very useful if query has recursive types) Docs - #69 -
DeferredResolver
now gotuserContext
as an argument. (breaking change: you need to provide a type parameter and one extra argument inresolve
for yourDeferredResolver
s. you you are not interested inuserContext
, you can just useAny
type) - Renamed Json support objects in order to make more concise import syntax (breaking change: you need to rename imports as well):
sangria.integration.CirceSupport
->sangria.integration.circe
sangria.integration.Json4sSupport
->sangria.integration.json4s
sangria.integration.PlayJsonSupport
->sangria.integration.playJson
sangria.integration.SprayJsonSupport
->sangria.integration.sprayJson
ResultMarshaller
andInputUnmarshaller
are moved in theintegration
package- Renamed execution
arguments
tovariables
in order to be consistent with the spec (breaking change: you need to rename this argument as well, if you are using named arguments) - Refactored variables and
InputUnmarshaller
. In order to avoid extra complexity it now does not have a dependent type. Instead it uses "type tagging" for scala map variables. It's a minor breaking change. If you are providing execution variables as a scala map, then you need to usemapVars
oremptyMapVars
which are defined inInputUnmarshaller
companion object (these functions do not wrapMap
- they only needed to ensure type constraints):Executor.execute(mySchema, query, variables = mapVars(Map("someId" -> "1000"))) // or Executor.execute(mySchema, query, variables = mapVars("someId" -> "1000"))
- #72 -
scala.util.Try
now can be returned fromresolve
in order to indicate a successful or failed result - #65 -
DeprecationTracker
should be called even if deprecation is in the interface type - #66 -
DeprecationTracker
should provide more contextual information (breaking change: the signature ofdeprecatedFieldUsed
is changed. It now provides much more contextual information, but you need to update the code that implements it) - #74 - Improved unicode handling (spec change)
- #67 - circe integration throws NoSuchElementException during execution
- #75 - Identical documents should be equal
- #73 - Verify input field uniqueness (spec change - new validation rule)
- Minor bugfixes
- #58 - Implement CirceJsonSupport in order to be able to integrate with Circe
- #53 - Add
map
inAction
- #53 - Ensure Ctx proper inheritance behavior
- #33 -
graphql
string context macro to create parsed document and verify query at compile time (big thanks to @dlreeves for implementing this feature). Here is an example how you can use it:
import sangria.macros._
val queryAst = graphql"""
query FetchSomeIDQuery {
human(id: "1000") {
name
}
}
"""
If there is a syntax error in the query, you will see it at the compile time.
-
#45 - Added
Long
scalar type -
#49 -
UpdateCtxAction
should also work for query types -
#50 - Sanity check - fields should have unique name within the same type definition
-
#31, #32 - More test coverage for "projections" and "deferred" features
-
#51 - Custom exception handler now should return
message
and list of additional filed -
The
interfaces
property syntax changed. In order to ensure type safety, improve type inference and allow type-class based relations betweenInterfaceType
andObjectType
you now need to use following syntax:val PersonType = ObjectType("Person", interfaces = interfaces[Unit, Person](NamedType, BeingType), fields = ...)
instead of old syntax
val PersonType = ObjectType[Unit, Person]("Person", interfaces = NamedType :: BeingType :: Nil, fields = ...) // or val PersonType = ObjectType[Unit, Person]("Person", interfaces = List(NamedType, BeingType), fields = ...)
-
Fields in
ObjectType
andInterfaceType
got small convenience method fields. You now can use it like this:val DogType = ObjectType("Dog", fields[Unit, Dog]( Field("name", OptionType(StringType), resolve = _.value.name), Field("barks", OptionType(BooleanType), resolve = _.value.barks)))
-
withPossibleTypes
was introduced onInterfaceType
andField
in order to provide a convenient way to the list of possible implementation types of interface -
Added convenience method
Executor.execute
-
Other minor improvements to make sangria-relay possible
- #44 - Add ability to add types explicitly in schema, for cases when they are not referenced anywhere else
Schema
now has additional argumentadditionalTypes
which can be used like this:Schema(HeroOnlyQuery, additionalTypes = Human :: Droid :: Nil)
- Spec changes - sew validation rules:
LoneAnonymousOperation
UniqueArgumentNames
UniqueFragmentNames
UniqueOperationNames
- groupId is changed to org.sangria-graphql
- Added missing query validation rules (#30 #29 #28 #27 #26 #25 #24 #23)
- #36 - Change semantics of
Projector
featureProjector
now gets all of the fields, not only fields marked withProjection
Projection
now allows to customize the nameNoProjection
allows to remove field from projections listProjector
allows to specify how deep it should look (the level arg)
Initial feature-complete release