Trivial first filesystem loader.

Assumes every schema is internally identified and knows what spec it
belongs to.

Successive commits will loosen those constraints, as well as make this
work for things which are even less Path-like (sigh,
importlib.abc.Traversable).

Author: Julian

docs/requirements.txt

@@ -39,7 +39,7 @@ pygments==2.17.2
     #   sphinx
 pygments-github-lexers==0.0.5
     # via -r docs/requirements.in
-referencing==0.31.0
+referencing==0.31.1
     # via referencing-loaders
 file:.#egg=referencing-loaders
     # via -r docs/requirements.in

pyproject.toml

@@ -38,7 +38,7 @@ classifiers = [
 ]
 dynamic = ["version"]
 dependencies = [
-  "referencing>=0.31.0",
+  "referencing>=0.31.1",
 ]
 
 [project.urls]

referencing/loaders.py

@@ -3,17 +3,32 @@
 """
 from __future__ import annotations
 
+from pathlib import Path
 from typing import TYPE_CHECKING, Any, Iterable
+import json
+import os
 
 from referencing import Resource
 
 if TYPE_CHECKING:
-    from pathlib import Path
+    from referencing.typing import URI
 
 
-def from_path(path: Path) -> Iterable[Resource[Any]]:
+def from_path(path: Path) -> Iterable[tuple[URI, Resource[Any]]]:
     """
     Load some resources recursively from a given directory path.
     """
-    for _, _, _ in path.walk():
-        yield Resource.opaque(None)
+    for root, _, files in _walk(path):
+        for file in files:
+            path = root / file
+            contents = json.loads(path.read_text())
+            yield path.as_uri(), Resource.from_contents(contents)
+
+
+def _walk(path: Path):
+    walk = getattr(path, "walk", None)
+    if walk is not None:
+        yield from walk()
+        return
+    for root, dirs, files in os.walk(path):
+        yield Path(root), dirs, files

referencing/tests/test_loaders.py

@@ -1,2 +1,51 @@
-def test_it_imports(tmp_path):
-    import referencing.loaders  # noqa: F401
+import json
+
+from referencing import Registry, loaders
+from referencing.jsonschema import EMPTY_REGISTRY
+
+
+def test_absolute_internally_identified(tmp_path):
+    root_path, root = tmp_path / "schema.json", {
+        "$schema": "https://json-schema.org/draft/2020-12/schema",
+        "$id": "http://example.com/",
+    }
+    son_path, son = tmp_path / "child/son.json", {
+        "$schema": "https://json-schema.org/draft/2019-09/schema",
+        "$id": "http://example.com/son",
+    }
+    daughter_path, daughter = tmp_path / "child/daughter.json", {
+        "$schema": "https://json-schema.org/draft/2020-12/schema",
+        "$id": "http://example.com/random/daughter",
+    }
+    grandchild_path, grandchild = tmp_path / "child/more/gc.json", {
+        "$schema": "https://json-schema.org/draft/2019-09/schema",
+        "$id": "http://example.com/also/a/grandchild",
+    }
+
+    tmp_path.joinpath("child/more").mkdir(parents=True)
+    root_path.write_text(json.dumps(root))
+    son_path.write_text(json.dumps(son))
+    daughter_path.write_text(json.dumps(daughter))
+    grandchild_path.write_text(json.dumps(grandchild))
+
+    expected = Registry().with_contents(
+        [
+            (root_path.as_uri(), root),
+            (son_path.as_uri(), son),
+            (daughter_path.as_uri(), daughter),
+            (grandchild_path.as_uri(), grandchild),
+        ],
+    )
+
+    resources = loaders.from_path(tmp_path)
+    registry = EMPTY_REGISTRY.with_resources(resources)
+    assert registry.crawl() == expected.crawl()
+
+
+def test_empty(tmp_path):
+    registry = EMPTY_REGISTRY.with_resources(loaders.from_path(tmp_path))
+    assert registry == EMPTY_REGISTRY
+
+
+def test_custom_loads(tmp_path):
+    pass

test-requirements.txt

@@ -14,7 +14,7 @@ pluggy==1.3.0
     # via pytest
 pytest==7.4.3
     # via -r test-requirements.in
-referencing==0.31.0
+referencing==0.31.1
     # via referencing-loaders
 file:.#egg=referencing-loaders
     # via -r test-requirements.in