CHANGELOG.md

Changelog

2.3.0 - 2024-05-30

Ruby Versions

• Ruby 2.5 and 2.6 are no longer supported.

Bug Fixes

• Remove base64 runtime dependency: #182
• Relax uuid format validation: #183

2.2.0 - 2024-03-02

Bug Fixes

• Support symbol keys when accessing original instance: https://github.com/davishmcclurg/json_schemer/commit/d52c130e9967919c6cf1c9dbc3f0babfb8b01cf8
• Support custom keywords in nested schemas: https://github.com/davishmcclurg/json_schemer/commit/93c85a5006981347c7e9a4c11b73c6bdb65d8ba2
• Stringify instance location for custom keywords: https://github.com/davishmcclurg/json_schemer/commit/513c99130b9e7986b09881e7efd3fb7143744754
• Reduce unhelpful error output in unevaluated keywords: #164
• Handle parse errors during schema validation: #171
• Follow refs when finding default property values: #175

Features

• Global configuration with Configuration object: #170
• Symbol key property defaults with insert_property_defaults: :symbol: https://github.com/davishmcclurg/json_schemer/commit/a72473dc84199107ddedc8998950e5b82273232a
• Consistent schema type support for schema validation methods: https://github.com/davishmcclurg/json_schemer/commit/bbcd0cea20cbaa61cf2bdae5f53840861cae54b8
• Validation option support for schema validation methods: https://github.com/davishmcclurg/json_schemer/commit/2eeef77de522f127619b7d0faa51e0d7e40977ad

2.1.1 - 2023-11-28

Bug Fixes

• Fix refs to/through keyword objects: #160
• Temporary fix for incorrect uri-reference format in OpenAPI 3.x: #161

2.1.0 - 2023-11-17

Bug Fixes

• Limit anyOf/oneOf discriminator to listed refs: #145
• Require discriminator propertyName property: #145
• Support Schema#ref in subschemas: #145
• Resolve JSON pointer refs using correct base URI: #147
• date format in OpenAPI 3.0: https://github.com/davishmcclurg/json_schemer/commit/69fe7a815ecf0cfb1c40ac402bf46a789c05e972

Features

• Custom error messages with x-error keyword and I18n: #149
• Custom content encodings and media types: #148

2.0.0 - 2023-08-20

For 2.0.0, much of the codebase was rewritten to simplify support for the two new JSON Schema draft versions (2019-09 and 2020-12). The major change is moving each keyword into its own class and organizing them into vocabularies. Output formats and annotations from the new drafts are also supported. The known breaking changes are listed below, but there may be others that haven't been identified.

Breaking Changes

• The default meta schema is now Draft 2020-12. Other meta schemas can be specified using meta_schema.
• Schemas use json-schemer://schema as the default base URI. Relative $id and $ref values are joined to the default base URI and are always absolute. For example, the schema { '$id' => 'foo', '$ref' => 'bar' } uses json-schemer://schema/foo as the base URI and passes json-schemer://schema/bar to the ref resolver. For relative refs, URI#path can be used in the ref resolver to access the relative portion, ie: URI('json-schemer://schema/bar').path => "/bar".
• Property validation hooks (before_property_validation and after_property_validation) run immediately before and after properties validation. Previously, before_property_validation ran before all "object" validations (dependencies, patternProperties, additionalProperties, etc) and after_property_validation was called after them.
• insert_property_defaults now inserts defaults in conditional subschemas when possible (if there's only one default or if there's only one unique default from a valid subtree).
• Error output
  • Special characters in schema_pointer are no longer percent encoded (eg, definitions/foo\"bar instead of /definitions/foo%22bar)
  • Keyword validation order changed so errors may be returned in a different order (eg, items errors before contains).
  • Array dependencies return "type": "dependencies" errors instead of "required" and point to the schema that contains the dependencies keyword.
  • not errors point to the schema that contains the not keyword (instead of the schema defined by the not keyword).
  • Custom keyword errors are now always wrapped in regular error hashes. Returned strings are used to set type:

    >> JSONSchemer.schema({ 'x' => 'y' }, :keywords => { 'x' => proc { false } }).validate({}).to_a
    => [{"data"=>{}, "data_pointer"=>"", "schema"=>{"x"=>"y"}, "schema_pointer"=>"", "root_schema"=>{"x"=>"y"}, "type"=>"x"}]
    >> JSONSchemer.schema({ 'x' => 'y' }, :keywords => { 'x' => proc { 'wrong!' } }).validate({}).to_a
    => [{"data"=>{}, "data_pointer"=>"", "schema"=>{"x"=>"y"}, "schema_pointer"=>"", "root_schema"=>{"x"=>"y"}, "type"=>"wrong!"}]
    

1.0.0 - 2023-05-26

Breaking Changes

• Ruby 2.4 is no longer supported.
• The default regexp_resolver is now ruby, which passes patterns directly to Regexp. The previous default, ecma, rewrites patterns to behave more like Javascript (ECMA-262) regular expressions:
  • Beginning of string: ^ -> \A
  • End of string: $ -> \z
  • Space: \s -> [\t\r\n\f\v\uFEFF\u2029\p{Zs}]
  • Non-space: \S -> [^\t\r\n\f\v\uFEFF\u2029\p{Zs}]
• Invalid ECMA-262 regular expressions raise JSONSchemer::InvalidEcmaRegexp when regexp_resolver is set to ecma.
• Embedded subschemas (ie, subschemas referenced by $id) can only be found under "known" keywords (eg, definitions). Previously, the entire schema object was scanned for $id.
• Empty fragments are now removed from $ref URIs before calling ref_resolver.
• Refs that are fragment-only JSON pointers with special characters must use the proper encoding (eg, "$ref": "#/definitions/some-%7Bid%7D").