chore: Split Registry

Signed-off-by: Dmitry Dygalo <dmitry@dygalo.dev>

Author: Stranger6667

CHANGELOG.md

@@ -2,6 +2,14 @@
 
 ## [Unreleased]
 
+### Breaking Changes
+
+- Registry construction now uses an explicit prepare step, and `with_registry` now borrows the prepared registry. `ValidationOptions::with_resource` and `ValidationOptions::with_resources` were removed in favor of building a `Registry` first. See the [Migration Guide](MIGRATION.md) for the details.
+
+### Performance
+
+- Avoid registry clones and document clones during validator construction. This improves real-world schema compilation by roughly 10-20% in internal benchmarks.
+
 ### Fixed
 
 - Incorrect handling of `duration` format when hours and seconds appear without minutes, or years and days without months.

MIGRATION.md

@@ -1,5 +1,74 @@
 # Migration Guide
 
+## Upgrading from 0.45.x to 0.46.0
+
+Registry construction is now explicit: add shared schemas first, then call
+`prepare()` to build a reusable registry. Validators no longer take ownership of
+that registry; pass it by reference with `with_registry(&registry)`.
+`ValidationOptions::with_resource` and `ValidationOptions::with_resources` were
+removed in favor of building a `Registry` first. For cases with multiple shared
+schemas, `extend([...])` is the batch form of `add(...)`.
+
+```rust
+// Old (0.45.x)
+use jsonschema::{Registry, Resource};
+
+// Inline shared schema
+let validator = jsonschema::options()
+    .with_resource(
+        "https://example.com/schema",
+        Resource::from_contents(shared_schema),
+    )
+    .build(&schema)?;
+
+// Multiple shared schemas
+let validator = jsonschema::options()
+    .with_resources([
+        (
+            "https://example.com/schema-1",
+            Resource::from_contents(schema_1),
+        ),
+        (
+            "https://example.com/schema-2",
+            Resource::from_contents(schema_2),
+        ),
+    ].into_iter())
+    .build(&schema)?;
+
+// Prebuilt registry
+let registry = Registry::try_from_resources([
+    (
+        "https://example.com/schema",
+        Resource::from_contents(shared_schema),
+    ),
+])?;
+let validator = jsonschema::options()
+    .with_registry(registry)
+    .build(&schema)?;
+
+// New (0.46.0)
+use jsonschema::Registry;
+
+// Shared registry + borrowed validator build
+let registry = Registry::new()
+    .add("https://example.com/schema", shared_schema)?
+    .prepare()?;
+let validator = jsonschema::options()
+    .with_registry(&registry)
+    .build(&schema)?;
+
+// Multiple shared schemas
+let registry = Registry::new()
+    .extend([
+        ("https://example.com/schema-1", schema_1),
+        ("https://example.com/schema-2", schema_2),
+    ])?
+    .prepare()?;
+let validator = jsonschema::options()
+    .with_registry(&registry)
+    .build(&schema)?;
+```
+
 ## Upgrading from 0.38.x to 0.39.0
 
 ### Custom keyword API simplified

crates/jsonschema-cli/src/main.rs

@@ -444,10 +444,10 @@ fn path_to_uri(path: &std::path::Path) -> String {
     result
 }
 
-fn options_for_schema(
+fn options_for_schema<'a>(
     schema_path: &Path,
     http_options: Option<&jsonschema::HttpOptions>,
-) -> Result<jsonschema::ValidationOptions, Box<dyn std::error::Error>> {
+) -> Result<jsonschema::ValidationOptions<'a>, Box<dyn std::error::Error>> {
     let base_uri = path_to_uri(schema_path);
     let base_uri = referencing::uri::from_str(&base_uri)?;
     let mut options = jsonschema::options().with_base_uri(base_uri);
@@ -695,16 +695,30 @@ fn run_bundle(args: BundleArgs) -> ExitCode {
         Err(error) => return fail_with_error(error),
     };
 
+    let mut registry = if let Some(http_opts) = http_options.as_ref() {
+        let retriever = match jsonschema::HttpRetriever::new(http_opts) {
+            Ok(retriever) => retriever,
+            Err(error) => return fail_with_error(error),
+        };
+        jsonschema::Registry::new().retriever(retriever)
+    } else {
+        jsonschema::Registry::new()
+    };
     for (uri, path) in &resources {
         let resource_json = match read_json(path) {
             Ok(value) => value,
             Err(error) => return fail_with_error(error),
         };
-        opts = opts.with_resource(
-            uri.as_str(),
-            referencing::Resource::from_contents(resource_json),
-        );
+        registry = match registry.add(uri, resource_json) {
+            Ok(registry) => registry,
+            Err(error) => return fail_with_error(error),
+        };
     }
+    let registry = match registry.prepare() {
+        Ok(registry) => registry,
+        Err(error) => return fail_with_error(error),
+    };
+    opts = opts.with_registry(&registry);
 
     match opts.bundle(&schema_json) {
         Ok(bundled) => {

crates/jsonschema-py/src/lib.rs

@@ -832,19 +832,19 @@ impl jsonschema::Keyword for CustomKeyword {
     }
 }
 
-fn make_options(
+fn make_options<'a>(
     draft: Option<u8>,
-    formats: Option<&Bound<'_, PyDict>>,
+    formats: Option<&Bound<'a, PyDict>>,
     validate_formats: Option<bool>,
     ignore_unknown_formats: Option<bool>,
-    retriever: Option<&Bound<'_, PyAny>>,
-    registry: Option<&registry::Registry>,
+    retriever: Option<&Bound<'a, PyAny>>,
+    registry: Option<&'a registry::Registry>,
     base_uri: Option<String>,
-    pattern_options: Option<&Bound<'_, PyAny>>,
-    email_options: Option<&Bound<'_, PyAny>>,
-    http_options: Option<&Bound<'_, PyAny>>,
-    keywords: Option<&Bound<'_, PyDict>>,
-) -> PyResult<jsonschema::ValidationOptions> {
+    pattern_options: Option<&Bound<'a, PyAny>>,
+    email_options: Option<&Bound<'a, PyAny>>,
+    http_options: Option<&Bound<'a, PyAny>>,
+    keywords: Option<&Bound<'a, PyDict>>,
+) -> PyResult<jsonschema::ValidationOptions<'a>> {
     let mut options = jsonschema::options();
     if let Some(raw_draft_version) = draft {
         options = options.with_draft(get_draft(raw_draft_version)?);
@@ -890,7 +890,7 @@ fn make_options(
         options = options.with_retriever(Retriever { func });
     }
     if let Some(registry) = registry {
-        options = options.with_registry(registry.inner.clone());
+        options = options.with_registry(&registry.inner);
     }
     if let Some(base_uri) = base_uri {
         options = options.with_base_uri(base_uri);
@@ -2021,7 +2021,7 @@ mod meta {
         let schema = crate::ser::to_value(schema)?;
         let result = if let Some(registry) = registry {
             jsonschema::meta::options()
-                .with_registry(registry.inner.clone())
+                .with_registry(&registry.inner)
                 .validate(&schema)
         } else {
             jsonschema::meta::validate(&schema)
@@ -2070,7 +2070,7 @@ mod meta {
         let schema = crate::ser::to_value(schema)?;
         let result = if let Some(registry) = registry {
             jsonschema::meta::options()
-                .with_registry(registry.inner.clone())
+                .with_registry(&registry.inner)
                 .validate(&schema)
         } else {
             jsonschema::meta::validate(&schema)

crates/jsonschema-py/src/registry.rs

@@ -1,12 +1,11 @@
-use jsonschema::Resource;
 use pyo3::{exceptions::PyValueError, prelude::*};
 
 use crate::{get_draft, retriever::into_retriever, to_value, Retriever};
 
 /// A registry of JSON Schema resources, each identified by their canonical URIs.
 #[pyclass]
 pub(crate) struct Registry {
-    pub(crate) inner: jsonschema::Registry,
+    pub(crate) inner: jsonschema::Registry<'static>,
 }
 
 #[pymethods]
@@ -19,30 +18,29 @@ impl Registry {
         draft: Option<u8>,
         retriever: Option<&Bound<'_, PyAny>>,
     ) -> PyResult<Self> {
-        let mut options = jsonschema::Registry::options();
+        let mut builder = jsonschema::Registry::new();
 
         if let Some(draft) = draft {
-            options = options.draft(get_draft(draft)?);
+            builder = builder.draft(get_draft(draft)?);
         }
 
         if let Some(retriever) = retriever {
             let func = into_retriever(retriever)?;
-            options = options.retriever(Retriever { func });
+            builder = builder.retriever(Retriever { func });
         }
 
-        let pairs = resources.try_iter()?.map(|item| {
+        for item in resources.try_iter()? {
             let pair = item?.unbind();
             let (key, value) = pair.extract::<(Bound<PyAny>, Bound<PyAny>)>(py)?;
             let uri = key.extract::<String>()?;
             let schema = to_value(&value)?;
-            let resource = Resource::from_contents(schema);
-            Ok((uri, resource))
-        });
-
-        let pairs: Result<Vec<_>, PyErr> = pairs.collect();
+            builder = builder
+                .add(uri, schema)
+                .map_err(|e| PyValueError::new_err(e.to_string()))?;
+        }
 
-        let registry = options
-            .build(pairs?)
+        let registry = builder
+            .prepare()
             .map_err(|e| PyValueError::new_err(e.to_string()))?;
 
         Ok(Registry { inner: registry })

crates/jsonschema-py/tests-py/test_bundle.py

@@ -43,6 +43,53 @@ def test_bundle_validates_identically():
     assert not validator.is_valid({"age": 30})
 
 
+def test_bundle_with_registry_and_explicit_draft4_legacy_id_root():
+    root = {
+        "id": "urn:root",
+        "type": "object",
+        "properties": {"value": {"$ref": "urn:string"}},
+        "required": ["value"],
+    }
+    registry = jsonschema_rs.Registry(
+        resources=[("urn:string", {"type": "string"})],
+        draft=jsonschema_rs.Draft4,
+    )
+
+    bundled = jsonschema_rs.bundle(root, registry=registry, draft=jsonschema_rs.Draft4)
+
+    assert bundled["properties"]["value"]["$ref"] == "urn:string"
+    assert "urn:string" in bundled["definitions"]
+
+
+def test_bundle_uses_call_retriever_when_inline_root_adds_external_ref():
+    def retrieve(uri: str):
+        if uri == "urn:external":
+            return {"type": "string"}
+        raise KeyError(f"Schema not found: {uri}")
+
+    root = {
+        "type": "object",
+        "properties": {"value": {"$ref": "urn:external"}},
+        "required": ["value"],
+    }
+    registry = jsonschema_rs.Registry(resources=[("urn:seed", {"type": "integer"})])
+
+    bundled = jsonschema_rs.bundle(root, registry=registry, retriever=retrieve)
+
+    assert bundled["properties"]["value"]["$ref"] == "urn:external"
+    assert "urn:external" in bundled["$defs"]
+
+
+def test_bundle_with_registry_accepts_equivalent_base_uri_with_empty_fragment():
+    root = {"$id": "urn:root", "$ref": "urn:shared"}
+    registry = jsonschema_rs.Registry(resources=[("urn:shared", {"type": "integer"})])
+
+    bundled = jsonschema_rs.bundle(root, registry=registry, base_uri="urn:root#")
+
+    assert bundled["$ref"] == "urn:shared"
+    assert bundled["$defs"]["urn:shared"]["type"] == "integer"
+
+
 def test_bundle_unresolvable_raises():
     with pytest.raises(jsonschema_rs.ReferencingError):
         jsonschema_rs.bundle({"$ref": "https://example.com/missing.json"})

crates/jsonschema-py/tests-py/test_registry.py

@@ -85,7 +85,6 @@ def test_top_level_functions_with_registry(function):
         assert list(function(schema, VALID_PERSON, registry=registry)) == []
         assert list(function(schema, INVALID_PERSON, registry=registry)) != []
 
-
 def test_validator_for_with_registry():
     registry = Registry(NESTED_RESOURCES)
     schema = {"$ref": "https://example.com/person.json"}
@@ -96,6 +95,21 @@ def test_validator_for_with_registry():
     assert not validator.is_valid(INVALID_PERSON)
 
 
+def test_validator_for_with_registry_and_explicit_draft4_legacy_id_root():
+    registry = Registry([("urn:string", {"type": "string"})], draft=Draft4)
+    schema = {
+        "id": "urn:root",
+        "type": "object",
+        "properties": {"value": {"$ref": "urn:string"}},
+        "required": ["value"],
+    }
+
+    validator = Draft4Validator(schema, registry=registry)
+
+    assert validator.is_valid({"value": "ok"})
+    assert not validator.is_valid({"value": 42})
+
+
 def test_registry_with_retriever_and_validation():
     def retrieve(uri: str):
         if uri == "https://example.com/dynamic.json":
@@ -118,6 +132,34 @@ def retrieve(uri: str):
     assert not dynamic_validator.is_valid("test")
 
 
+def test_validator_for_uses_call_retriever_when_inline_root_adds_external_ref():
+    def retrieve(uri: str):
+        if uri == "urn:external":
+            return {"type": "string"}
+        raise KeyError(f"Schema not found: {uri}")
+
+    registry = Registry([("urn:seed", {"type": "integer"})])
+    schema = {
+        "type": "object",
+        "properties": {"value": {"$ref": "urn:external"}},
+        "required": ["value"],
+    }
+
+    validator = validator_for(schema, registry=registry, retriever=retrieve)
+    assert validator.is_valid({"value": "ok"})
+    assert not validator.is_valid({"value": 42})
+
+
+def test_validator_for_with_registry_accepts_equivalent_base_uri_with_empty_fragment():
+    registry = Registry([("urn:shared", {"type": "integer"})])
+    schema = {"$id": "urn:root", "$ref": "urn:shared"}
+
+    validator = validator_for(schema, registry=registry, base_uri="urn:root#")
+
+    assert validator.is_valid(1)
+    assert not validator.is_valid("x")
+
+
 def test_registry_error_propagation():
     registry = Registry(NESTED_RESOURCES)
 

crates/jsonschema-rb/spec/bundle_spec.rb

@@ -43,6 +43,23 @@
     expect(validator.valid?({ "age" => 30 })).to be false
   end
 
+  it "bundles inline legacy-id root with registry and explicit draft4" do
+    root = {
+      "id" => "urn:root",
+      "type" => "object",
+      "properties" => { "value" => { "$ref" => "urn:string" } },
+      "required" => ["value"]
+    }
+    registry = JSONSchema::Registry.new(
+      [["urn:string", { "type" => "string" }]],
+      draft: :draft4
+    )
+
+    bundled = JSONSchema.bundle(root, registry: registry, draft: :draft4)
+    expect(bundled.dig("properties", "value", "$ref")).to eq("urn:string")
+    expect(bundled.dig("definitions", "urn:string")).not_to be_nil
+  end
+
   it "raises when a $ref cannot be resolved" do
     expect do
       JSONSchema.bundle({ "$ref" => "https://example.com/missing.json" })