Add JSON Schema draft 2019-09 and 2020-12 support

CHANGELOG.md

@@ -1,5 +1,28 @@
 # CHANGELOG
 
+## Unreleased
+
+* Add partial support for JSON Schema draft 2019-09 and draft 2020-12. New
+  dialect modules `jesse_validator_draft2019_09` and
+  `jesse_validator_draft2020_12`, dispatched from the schema's `$schema` URI
+  (`https://json-schema.org/draft/2019-09/schema` and
+  `.../2020-12/schema`, with or without a trailing `#`). Implemented keywords:
+  `dependentRequired`, `dependentSchemas`, `if`/`then`/`else`,
+  `minContains`/`maxContains`, `$defs`, local `$anchor` resolution, `$ref`
+  evaluated alongside sibling keywords, and
+  `unevaluatedProperties`/`unevaluatedItems` (with the full annotation
+  model — adjacent keywords and successful in-place applicators contribute,
+  cousins/uncles do not). Draft 2020-12 additionally handles the `prefixItems`
+  rename (tuple validation) and `items` as the after-`prefixItems` applicator.
+  `format` is annotation-only (non-asserting), per the dialect default.
+  Not yet implemented — `$recursiveRef` (2019-09) and `$dynamicRef` (2020-12) —
+  raise a `keyword_not_supported` schema error instead of being silently
+  ignored, so they can never false-accept invalid data.
+  Validated against the official JSON-Schema-Test-Suite: 927/1003 individual
+  draft 2019-09 tests and 935/997 draft 2020-12 tests pass; the remainder are
+  for `$recursiveRef`/`$dynamicRef`, remote-schema fetching, and in-document
+  `$id` scoping.
+
 ## 1.5.6
 
 * Improving the error messages from jesse when using oneOf/anyOf

Makefile

@@ -116,6 +116,14 @@ symlinks: test/JSON-Schema-Test-Suite/tests
 	ln -sf ../../test/JSON-Schema-Test-Suite/tests/draft6 standard && \
 	ln -sf ../../test/JSON-Schema-Test-Suite/remotes remotes
 
+	cd test/jesse_tests_draft2019_09_SUITE_data && \
+	ln -sf ../../test/JSON-Schema-Test-Suite/tests/draft2019-09 standard && \
+	ln -sf ../../test/JSON-Schema-Test-Suite/remotes remotes
+
+	cd test/jesse_tests_draft2020_12_SUITE_data && \
+	ln -sf ../../test/JSON-Schema-Test-Suite/tests/draft2020-12 standard && \
+	ln -sf ../../test/JSON-Schema-Test-Suite/remotes remotes
+
 # https://github.com/erlang/rebar3/issues/2903
 ./rebar3.OTP18:
 	$(CURL) -qfsSL -o $@ https://github.com/erlang/rebar3/releases/download/3.15.3/rebar3 && $(CHMOD) +x $@

README.md

@@ -8,6 +8,12 @@ jesse implements the following specifications:
 * [Draft 03](http://tools.ietf.org/html/draft-zyp-json-schema-03)
 * [Draft 04](http://tools.ietf.org/html/draft-zyp-json-schema-04)
 * [Draft 06](https://datatracker.ietf.org/doc/html/draft-wright-json-schema-00)
+* [Draft 2019-09](https://json-schema.org/draft/2019-09/json-schema-core.html)
+  (partial: `$recursiveRef` and remote-schema fetching are not yet supported
+  and raise an error rather than being silently ignored)
+* [Draft 2020-12](https://json-schema.org/draft/2020-12/json-schema-core.html)
+  (partial: `$dynamicRef` and remote-schema fetching are not yet supported
+  and raise an error rather than being silently ignored)
 
 Install from git or https://hex.pm/packages/jesse .
 
@@ -241,9 +247,14 @@ Maps example
 
 ## JSON Schema versions
 
-jesse currently supports JSON Schema draft3, draft4 and draft6. To decide which
+jesse currently supports JSON Schema draft3, draft4, draft6 and (partially)
+draft 2019-09 and draft 2020-12. To decide which
 validator to use jesse tries to read `$schema` property from the given schema,
 and checks if it's a supported one, otherwise it will return an error.
+For draft 2019-09/2020-12, keywords that the dialect defines but jesse does not
+yet implement (`$recursiveRef`, `$dynamicRef`) raise a `keyword_not_supported`
+schema error instead of being silently ignored, so a schema relying on them can
+never quietly accept data those keywords would reject.
 If `$schema` property isn't provided in the given schema, jesse will use the
 default validator (currently the validator for draft3).
 

rebar.config

@@ -54,7 +54,12 @@
                   , {elvis_style, no_behavior_info}
                   , { elvis_style
                     , module_naming_convention
-                    , #{ regex => "^([a-z][a-z0-9]*_?)*(_SUITE)?$"
+                      %% Allow purely-numeric underscore-delimited segments
+                      %% (e.g. the "_09" in jesse_validator_draft2019_09) so
+                      %% draft-year_month dialect modules can be named after
+                      %% their metaschema version.
+                    , #{ regex =>
+                           "^([a-z][a-z0-9]*)(_[a-z0-9]+)*(_SUITE)?$"
                        , ignore => []
                        }
                     }

src/jesse.erl

@@ -35,6 +35,7 @@
         , validate/3
         , validate_with_schema/2
         , validate_with_schema/3
+        , supported_dialect/1
         ]).
 
 -export_type([ allowed_errors/0
@@ -247,6 +248,15 @@ validate_with_schema(Schema, Data, Options) ->
     throw:Error -> {error, Error}
   end.
 
+%% @doc Whether the given `$schema' dialect URI is one jesse can validate
+%% against. Useful for rejecting a schema that declares an unsupported dialect
+%% at registration time, instead of accepting it and failing every validation
+%% at run time. Draft 2019-09/2020-12 URIs are accepted with or without a
+%% trailing `#'.
+-spec supported_dialect(SchemaURI :: binary()) -> boolean().
+supported_dialect(SchemaURI) ->
+  jesse_schema_validator:is_supported_dialect(SchemaURI).
+
 %%% Internal functions
 %% @doc Wraps up calls to a third party json parser.
 %% @private

src/jesse_lib.erl

@@ -266,6 +266,10 @@ compare_properties(Value1, Value2) ->
 get_schema_id_key(Schema) ->
   case jesse_json_path:value(?SCHEMA, Schema, ?json_schema_draft6) of
     ?json_schema_draft6 -> ?ID;
+    ?json_schema_draft2019_09 -> ?ID;
+    <<"https://json-schema.org/draft/2019-09/schema#">> -> ?ID;
+    ?json_schema_draft2020_12 -> ?ID;
+    <<"https://json-schema.org/draft/2020-12/schema#">> -> ?ID;
                       _ -> ?ID_OLD
   end.
 

src/jesse_schema_validator.erl

@@ -26,6 +26,7 @@
 %% API
 -export([ validate/3
         , validate_with_state/3
+        , is_supported_dialect/1
         ]).
 
 %% Includes
@@ -57,6 +58,23 @@ validate_with_state(JsonSchema, Value, State) ->
   SchemaVer = get_schema_ver(JsonSchema, State),
   select_and_run_validator(SchemaVer, JsonSchema, Value, State).
 
+%% @doc Whether a schema's `$schema' dialect URI is one jesse can validate
+%% against. Intended for callers that want to reject a schema declaring an
+%% unsupported dialect up-front (e.g. at registration) instead of having every
+%% validation fail at run time. Draft 2019-09/2020-12 URIs are accepted with or
+%% without a trailing `#'.
+-spec is_supported_dialect(SchemaURI :: binary()) -> boolean().
+is_supported_dialect(?json_schema_draft3) -> true;
+is_supported_dialect(?json_schema_draft4) -> true;
+is_supported_dialect(?json_schema_draft6) -> true;
+is_supported_dialect(SchemaURI) when is_binary(SchemaURI) ->
+  case normalize_schema_ver(SchemaURI) of
+    ?json_schema_draft2019_09 -> true;
+    ?json_schema_draft2020_12 -> true;
+    _ -> false
+  end;
+is_supported_dialect(_) -> false.
+
 %%% Internal functions
 %% @doc Returns "$schema" property from `JsonSchema' if it is present,
 %% otherwise the default schema version from `State' is returned.
@@ -94,5 +112,33 @@ select_and_run_validator(?json_schema_draft6, JsonSchema, Value, State) ->
                                       , jesse_json_path:unwrap_value(JsonSchema)
                                       , State
                                       );
-select_and_run_validator(SchemaURI, _JsonSchema, _Value, State) ->
-  jesse_error:handle_schema_invalid({?schema_unsupported, SchemaURI}, State).
+select_and_run_validator(SchemaURI, JsonSchema, Value, State) ->
+  case normalize_schema_ver(SchemaURI) of
+    ?json_schema_draft2019_09 ->
+      jesse_validator_draft2019_09:check_value(
+        Value, jesse_json_path:unwrap_value(JsonSchema), State);
+    ?json_schema_draft2020_12 ->
+      jesse_validator_draft2020_12:check_value(
+        Value, jesse_json_path:unwrap_value(JsonSchema), State);
+    _ ->
+      jesse_error:handle_schema_invalid({?schema_unsupported, SchemaURI}, State)
+  end.
+
+%% @doc Normalize a "$schema" URI so that draft 2019-09/2020-12 schemas dispatch
+%% regardless of a trailing "#" fragment or http/https scheme. Draft 3/4/6 are
+%% matched verbatim by the clauses above and never reach here.
+%% @private
+normalize_schema_ver(SchemaURI) when is_binary(SchemaURI) ->
+  Stripped =
+    case SchemaURI of
+      <<Base:(byte_size(SchemaURI) - 1)/binary, $#>> -> Base;
+      _ -> SchemaURI
+    end,
+  case Stripped of
+    <<"http://json-schema.org/", Rest/binary>> ->
+      <<"https://json-schema.org/", Rest/binary>>;
+    _ ->
+      Stripped
+  end;
+normalize_schema_ver(SchemaURI) ->
+  SchemaURI.

src/jesse_schema_validator.hrl

@@ -82,6 +82,26 @@
 -define(MAXPROPERTIES,        <<"maxProperties">>).
 -define(MINPROPERTIES,        <<"minProperties">>).
 
+%% Keywords introduced in draft 2019-09 / 2020-12
+-define(DEFS,                 <<"$defs">>).
+-define(ANCHOR,               <<"$anchor">>).
+-define(RECURSIVEANCHOR,      <<"$recursiveAnchor">>).
+-define(RECURSIVEREF,         <<"$recursiveRef">>).
+-define(DYNAMICANCHOR,        <<"$dynamicAnchor">>).
+-define(DYNAMICREF,           <<"$dynamicRef">>).
+-define(VOCABULARY,           <<"$vocabulary">>).
+-define(COMMENT,              <<"$comment">>).
+-define(DEPENDENTREQUIRED,    <<"dependentRequired">>).
+-define(DEPENDENTSCHEMAS,     <<"dependentSchemas">>).
+-define(IF,                   <<"if">>).
+-define(THEN,                 <<"then">>).
+-define(ELSE,                 <<"else">>).
+-define(MINCONTAINS,          <<"minContains">>).
+-define(MAXCONTAINS,          <<"maxContains">>).
+-define(PREFIXITEMS,          <<"prefixItems">>).
+-define(UNEVALUATEDPROPERTIES, <<"unevaluatedProperties">>).
+-define(UNEVALUATEDITEMS,     <<"unevaluatedItems">>).
+
 %% Constant definitions for Json types
 -define(ANY,                  <<"any">>).
 -define(ARRAY,                <<"array">>).
@@ -96,6 +116,14 @@
 -define(json_schema_draft3, <<"http://json-schema.org/draft-03/schema#">>).
 -define(json_schema_draft4, <<"http://json-schema.org/draft-04/schema#">>).
 -define(json_schema_draft6, <<"http://json-schema.org/draft-06/schema#">>).
+%% Draft 2019-09 and 2020-12 canonical metaschema URIs omit the trailing "#".
+%% Incoming "$schema" values are normalized (trailing "#" stripped, scheme
+%% coerced to https) before being matched against these, so both the
+%% fragment and non-fragment forms dispatch to the right dialect.
+-define(json_schema_draft2019_09,
+        <<"https://json-schema.org/draft/2019-09/schema">>).
+-define(json_schema_draft2020_12,
+        <<"https://json-schema.org/draft/2020-12/schema">>).
 -define(default_schema_ver, ?json_schema_draft6).
 -define(default_schema_loader_fun, fun jesse_database:load_uri/1).
 -define(default_error_handler_fun, fun jesse_error:default_error_handler/3).
@@ -107,6 +135,10 @@
 -define(schema_invalid,              'schema_invalid').
 -define(schema_not_found,            'schema_not_found').
 -define(schema_unsupported,          'schema_unsupported').
+%% Raised when a schema uses a keyword that the dialect defines but jesse does
+%% not yet implement. Surfacing it (instead of silently ignoring the keyword)
+%% prevents dangerous false-acceptance of data the keyword would have rejected.
+-define(keyword_not_supported,       'keyword_not_supported').
 -define(wrong_all_of_schema_array,   'wrong_all_of_schema_array').
 -define(wrong_any_of_schema_array,   'wrong_any_of_schema_array').
 -define(wrong_max_properties,        'wrong_max_properties').