artemis

• Fixed ignore_for_file documentation.

• Added ignore_for_file option to ignore linter rules on generated files.

• improved canonical types handling

• Package updates

• Fixed toJson() doesn't remove "kw$" prefix

• test fix

• analyzer and linter warnings fix

• uppercase keyword fix

• pubspec fix

• allow multiple operations per file

• performance improvements - scan schema for canonical types only once

• enum name pascal casing.

• Add deprecated annotations in fields.

• Build type name recursively, considering casing changes.

• Mass package update

• Centralize naming transformations; make types PascalCase and fields camelCase.

• Recursively consider input lists.

• Do not throw on unused scalars.

MAJOR BREAKING CHANGE

• all starting underscores are replaced with $
• __typename field replaced with $$typename
• enums are named according to Dart spec
• fields similar to Dart keywords are prefixed with kw$

• Check for more error causes and throw, explaining the error.

MAJOR BREAKING CHANGE

We've found a regression on 6.1.0-beta.1, which sends Enums as camelCase to the server, when they should be sent as SCREAMING_SNAKE_CASE.

• Reverts 6.1.0-beta.1.

• Improve actions and check pipeline output.

• Short-circuit input object generation on recursive detection

MAJOR BREAKING CHANGE

• Convert enum casing to camel case.

• Convert ClassProperty annotation item to List<String>.

• Duplication bug fix

• Added the exception for the case when fragment_glob leads to query files fragments ignore.

• Adapt Artemis to subscriptions and create an example

• Fix for the interfaces which uses fragments from fragments_glob

• Hide build logs under --verbose flag

• Properly consider "sub-fragments" on class generation.

• Fix generation of custom scalars and its functions.

• Fix invalid reference to class on Query generations.

MAJOR BREAKING CHANGE

• Generate canonical objects (enums and input objects) with their original names on GraphQL. Fragments are also generated with their own names (plus the Mixin prefix, for now).
• Make it possible to select a naming scheme to be used for generate the class names. pathedWithTypes is the default for retrocompatibility, where the names of previous types are used as prefix of the next class. This can generate duplication on certain schemas. With pathedWithFields, the names of previous fields are used as prefix of the next class and with simple, only the actual GraphQL class nameis considered. See discussion on [#90][pr-90] and [#96][pr-96] for more information.

• Add .graphql. to outputted files path, in a non-breaking change way: a "forwarder" file will be generated to make it retro-compatible when a configurated output doesn't end with .graphql.dart.

• Update CI to include beta branch.

• Update examples to match latest changes.

• Use default names for query/mutation root when SDL does not declare schema.

• Fix generation of recursive input objects introduced by 5.0.0.

MAJOR BREAKING CHANGE

In this version we moved from json to graphql (SDL) schema parsing. This allowed us to get rid off ±1200 lines of code which makes the
project support much easier. The test files with schema definitions became more clear and human readable.

If you already have your schema in SDL format, just point to it in build.yaml. If not, use this [snippet][introspection-to-sdl-snippet] (from [this Apollo article][apollo-3-ways-schema]) or online helpers like [this one][introspection-to-sdl-online] to convert from one to another.

• Only add unknownEnumValue on non-list enums
• Consider all classes to include reference to meta package

• Look at mutation root when generating a mutation

MAJOR BREAKING CHANGE

This version completely refactors how Artemis generate code (by finally using the implementation of visitor pattern provided by gql). On top of that, I've decided to do other major breaking changes to make code cleaner and more maintainable. Listed:

• add_query_prefix doesn't exist anymore (it's now the default to generate classes with its "path" from the query), e.g., this query's city field will be typed as CityName$QueryRoot$User$Address$City:

  query city_name {
    user {
      address {
        city {
          name
        }
      }
    }
  }
  

  This change was also done to tip users to NOT use those generated queries directly on their code, to avoid coupling them to your business logic.
• custom_parser_import was moved to inside a ScalarMap, and use_custom_parser was removed.
• resolve_type_field option was renamed to type_name_field, as __typename is the correct field name (by GraphQL spec).
• Classes generated for mutation will have a Mutation suffix, as queries already have Query suffix.
• Change pre-generation data classes constructors to named parameters, so if you're using GraphQLQueryBuilder.onBuild, it will break.

And also:

• Add more logs and errors while generating code, to help debugging.
• Add more/refactor tests.
• Add a GitHub example.

TODO:

• [ ] re-add more logs
• [ ] clean options (?)
• [ ] prefix every class with $ (?)
• [ ] refactor class naming variables
• [ ] review readme and changelog

• Fix unknown enum: add prefix

• Make enums loose. When unknown values are provided into an enum, it will fall back to a custom ARTEMIS_UNKNOWN value avoiding breaking/crashing the client.

• Allow to dispose ArtemisClient underlining http client when possible

• BREAKING: Marks non nullable input field as @required [#68][pr-68]

• Make lists as input objects work again

• Display error on types not found on schema

• Share fragments between queries and schemas (see fragments_glob) [#65][pr-65]

• Add "Articles and videos" category on README

• Add missing prefix to generated enums

• Bump equatable/gql suite, refine GitHub actions

• Bump json_serializable/json_annotation

• Properly consider Union types on generation

• Generate fragments as mixins

• Fix GitHub actions deploy pipeline
• Make sure artemis depends on json_annotation