Merge pull request #305 from python-jsonschema/explicit-base-uri

Add support for passing an explicit base URI

Author: sirosen

docs/usage.rst

@@ -40,8 +40,22 @@ Detailed helptext is always available interactively via
        the error and exit. Use ``--traceback-mode full`` to request the full traceback
        be printed, for debugging and troubleshooting.
 
-Other Schema Options
---------------------
+Environment Variables
+---------------------
+
+The following environment variables are supported.
+
+.. list-table:: Environment Variables
+   :widths: 15 30
+   :header-rows: 1
+
+   * - Name
+     - Description
+   * - ``NO_COLOR``
+     - Set ``NO_COLOR=1`` to explicitly turn off colorized output.
+
+Schema Selection Options
+------------------------
 
 No matter what usage form is used, a schema must be specified.
 
@@ -113,68 +127,6 @@ The following options control caching behaviors.
      - The name to use for caching a remote schema.
        Defaults to using the last slash-delimited part of the URI.
 
-Environment Variables
----------------------
-
-The following environment variables are supported.
-
-.. list-table:: Environment Variables
-   :widths: 15 30
-   :header-rows: 1
-
-   * - Name
-     - Description
-   * - ``NO_COLOR``
-     - Set ``NO_COLOR=1`` to explicitly turn off colorized output.
-
-Parsing Options
----------------
-
-``--default-filetype``
-~~~~~~~~~~~~~~~~~~~~~~
-
-The default filetype to assume on instance files when they are detected neither
-as JSON nor as YAML.
-
-For example, pass ``--default-filetype yaml`` to instruct that files which have
-no extension should be treated as YAML.
-
-By default, this is not set and files without a detected type of JSON or YAML
-will fail.
-
-``--data-transform``
-~~~~~~~~~~~~~~~~~~~~
-
-``--data-transform`` applies a transformation to instancefiles before they are
-checked. The following transforms are supported:
-
-- ``azure-pipelines``:
-    "Unpack" compile-time expressions for Azure Pipelines files, skipping them
-    for the purposes of validation. This transformation is based on Microsoft's
-    lanaguage-server for VSCode and how it handles expressions
-
-- ``gitlab-ci``:
-    Handle ``!reference`` tags in YAML data for gitlab-ci files. This transform
-    has no effect if the data is not being loaded from YAML, and it does not
-    interpret ``!reference`` usages -- it only expands them to lists of strings
-    to pass schema validation
-
-``--fill-defaults``
--------------------
-
-JSON Schema specifies the ``"default"`` keyword as potentially meaningful for
-consumers of schemas, but not for validators. Therefore, the default behavior
-for ``check-jsonschema`` is to ignore ``"default"``.
-
-``--fill-defaults`` changes this behavior, filling in ``"default"`` values
-whenever they are encountered prior to validation.
-
-.. warning::
-
-    There are many schemas which make the meaning of ``"default"`` unclear.
-    In particular, the behavior of ``check-jsonschema`` is undefined when multiple
-    defaults are specified via ``anyOf``, ``oneOf``, or other forms of polymorphism.
-
 "format" Validation Options
 ---------------------------
 
@@ -253,3 +205,61 @@ follows:
        always passes. Otherwise, check validity in the python engine.
    * - python
      - Require the regex to be valid in python regex syntax.
+
+Other Options
+--------------
+
+``--default-filetype``
+~~~~~~~~~~~~~~~~~~~~~~
+
+The default filetype to assume on instance files when they are detected neither
+as JSON nor as YAML.
+
+For example, pass ``--default-filetype yaml`` to instruct that files which have
+no extension should be treated as YAML.
+
+By default, this is not set and files without a detected type of JSON or YAML
+will fail.
+
+``--data-transform``
+~~~~~~~~~~~~~~~~~~~~
+
+``--data-transform`` applies a transformation to instancefiles before they are
+checked. The following transforms are supported:
+
+- ``azure-pipelines``:
+    "Unpack" compile-time expressions for Azure Pipelines files, skipping them
+    for the purposes of validation. This transformation is based on Microsoft's
+    lanaguage-server for VSCode and how it handles expressions
+
+- ``gitlab-ci``:
+    Handle ``!reference`` tags in YAML data for gitlab-ci files. This transform
+    has no effect if the data is not being loaded from YAML, and it does not
+    interpret ``!reference`` usages -- it only expands them to lists of strings
+    to pass schema validation
+
+``--fill-defaults``
+~~~~~~~~~~~~~~~~~~~
+
+JSON Schema specifies the ``"default"`` keyword as potentially meaningful for
+consumers of schemas, but not for validators. Therefore, the default behavior
+for ``check-jsonschema`` is to ignore ``"default"``.
+
+``--fill-defaults`` changes this behavior, filling in ``"default"`` values
+whenever they are encountered prior to validation.
+
+.. warning::
+
+    There are many schemas which make the meaning of ``"default"`` unclear.
+    In particular, the behavior of ``check-jsonschema`` is undefined when multiple
+    defaults are specified via ``anyOf``, ``oneOf``, or other forms of polymorphism.
+
+``--base-uri``
+~~~~~~~~~~~~~~
+
+``check-jsonschema`` defaults to using the ``"$id"`` of the schema as the base
+URI for ``$ref`` resolution, falling back to the retrieval URI if ``"$id"`` is
+not set.
+
+``--base-uri`` overrides this behavior, setting a custom base URI for ``$ref``
+resolution.

src/check_jsonschema/cli/main_command.py

@@ -94,6 +94,14 @@ def pretty_helptext_list(values: list[str] | tuple[str, ...]) -> str:
         "it will be downloaded and cached locally based on mtime."
     ),
 )
+@click.option(
+    "--base-uri",
+    help=(
+        "Override the base URI for the schema. The default behavior is to "
+        "follow the behavior specified by the JSON Schema spec, which is to "
+        "prefer an explicit '$id' and failover to the retrieval URI."
+    ),
+)
 @click.option(
     "--builtin-schema",
     help="The name of an internal schema to use for '--schemafile'",
@@ -212,6 +220,7 @@ def main(
     *,
     schemafile: str | None,
     builtin_schema: str | None,
+    base_uri: str | None,
     check_metaschema: bool,
     no_cache: bool,
     cache_filename: str | None,
@@ -230,6 +239,7 @@ def main(
     args = ParseResult()
 
     args.set_schema(schemafile, builtin_schema, check_metaschema)
+    args.base_uri = base_uri
     args.instancefiles = instancefiles
 
     normalized_disable_formats: tuple[str, ...] = tuple(
@@ -264,13 +274,18 @@ def main(
 
 def build_schema_loader(args: ParseResult) -> SchemaLoaderBase:
     if args.schema_mode == SchemaLoadingMode.metaschema:
-        return MetaSchemaLoader()
+        return MetaSchemaLoader(base_uri=args.base_uri)
     elif args.schema_mode == SchemaLoadingMode.builtin:
         assert args.schema_path is not None
-        return BuiltinSchemaLoader(args.schema_path)
+        return BuiltinSchemaLoader(args.schema_path, base_uri=args.base_uri)
     elif args.schema_mode == SchemaLoadingMode.filepath:
         assert args.schema_path is not None
-        return SchemaLoader(args.schema_path, args.cache_filename, args.disable_cache)
+        return SchemaLoader(
+            args.schema_path,
+            args.cache_filename,
+            args.disable_cache,
+            base_uri=args.base_uri,
+        )
     else:
         raise NotImplementedError("no valid schema option provided")
 

src/check_jsonschema/cli/parse_result.py

@@ -19,6 +19,7 @@ def __init__(self) -> None:
         # primary options: schema + instances
         self.schema_mode: SchemaLoadingMode = SchemaLoadingMode.filepath
         self.schema_path: str | None = None
+        self.base_uri: str | None = None
         self.instancefiles: tuple[str, ...] = ()
         # cache controls
         self.disable_cache: bool = False

src/check_jsonschema/schema_loader/main.py

@@ -61,11 +61,13 @@ def __init__(
         schemafile: str,
         cache_filename: str | None = None,
         disable_cache: bool = False,
+        base_uri: str | None = None,
     ) -> None:
         # record input parameters (these are not to be modified)
         self.schemafile = schemafile
         self.cache_filename = cache_filename
         self.disable_cache = disable_cache
+        self.base_uri = base_uri
 
         # if the schema location is a URL, which may include a file:// URL, parse it
         self.url_info = None
@@ -104,7 +106,10 @@ def get_schema_retrieval_uri(self) -> str | None:
         return self.reader.get_retrieval_uri()
 
     def get_schema(self) -> dict[str, t.Any]:
-        return self.reader.read_schema()
+        data = self.reader.read_schema()
+        if self.base_uri is not None:
+            data["$id"] = self.base_uri
+        return data
 
     def get_validator(
         self,
@@ -145,18 +150,29 @@ def get_validator(
 
 
 class BuiltinSchemaLoader(SchemaLoader):
-    def __init__(self, schema_name: str) -> None:
+    def __init__(self, schema_name: str, base_uri: str | None = None) -> None:
         self.schema_name = schema_name
+        self.base_uri = base_uri
         self._parsers = ParserSet()
 
     def get_schema_retrieval_uri(self) -> str | None:
         return None
 
     def get_schema(self) -> dict[str, t.Any]:
-        return get_builtin_schema(self.schema_name)
+        data = get_builtin_schema(self.schema_name)
+        if self.base_uri is not None:
+            data["$id"] = self.base_uri
+        return data
 
 
 class MetaSchemaLoader(SchemaLoaderBase):
+    def __init__(self, base_uri: str | None = None) -> None:
+        if base_uri is not None:
+            raise NotImplementedError(
+                "'--base-uri' was used with '--metaschema'. "
+                "This combination is not supported."
+            )
+
     def get_validator(
         self,
         path: pathlib.Path,

tests/acceptance/test_remote_ref_resolution.py

@@ -141,3 +141,43 @@ def test_ref_resolution_does_not_callout_for_absolute_ref_to_retrieval_uri(
         assert result.exit_code == 0, output
     else:
         assert result.exit_code == 1, output
+
+
+# this test ensures that `$id` is overwritten when `--base-uri` is used
+@pytest.mark.parametrize("check_passes", (True, False))
+def test_ref_resolution_with_custom_base_uri(run_line, tmp_path, check_passes):
+    retrieval_uri = "https://example.org/retrieval-and-in-schema-only/schemas/main"
+    explicit_base_uri = "https://example.org/schemas/main"
+    main_schema = {
+        "$id": retrieval_uri,
+        "$schema": "http://json-schema.org/draft-07/schema",
+        "properties": {
+            "title": {"$ref": "./title_schema.json"},
+        },
+        "additionalProperties": False,
+    }
+    title_schema = {"type": "string"}
+
+    responses.add("GET", retrieval_uri, json=main_schema)
+    responses.add(
+        "GET", "https://example.org/schemas/title_schema.json", json=title_schema
+    )
+
+    instance_path = tmp_path / "instance.json"
+    instance_path.write_text(json.dumps({"title": "doc one" if check_passes else 2}))
+
+    result = run_line(
+        [
+            "check-jsonschema",
+            "--schemafile",
+            retrieval_uri,
+            "--base-uri",
+            explicit_base_uri,
+            str(instance_path),
+        ]
+    )
+    output = f"\nstdout:\n{result.stdout}\n\nstderr:\n{result.stderr}"
+    if check_passes:
+        assert result.exit_code == 0, output
+    else:
+        assert result.exit_code == 1, output