Another placeholder benchmark for future optimization.

Author: Julian

jsonschema/benchmarks/useless_applicator_schemas.py

@@ -0,0 +1,106 @@
+
+"""
+A benchmark for validation of applicators containing lots of useless schemas.
+
+Signals a small possible optimization to remove all such schemas ahead of time.
+"""
+
+from pyperf import Runner
+
+from jsonschema import Draft202012Validator as Validator
+
+NUM_USELESS = 100000
+
+subschema = {"const": 37}
+
+valid = 37
+invalid = 12
+
+baseline = Validator(subschema)
+
+
+# These should be indistinguishable from just `subschema`
+by_name = {
+    "single subschema": {
+        "anyOf": Validator({"anyOf": [subschema]}),
+        "allOf": Validator({"allOf": [subschema]}),
+        "oneOf": Validator({"oneOf": [subschema]}),
+    },
+    "redundant subschemas": {
+        "anyOf": Validator({"anyOf": [subschema] * NUM_USELESS}),
+        "allOf": Validator({"allOf": [subschema] * NUM_USELESS}),
+    },
+    "useless successful subschemas (beginning)": {
+        "anyOf": Validator({"anyOf": [subschema, *[True] * NUM_USELESS]}),
+        "allOf": Validator({"allOf": [subschema, *[True] * NUM_USELESS]}),
+    },
+    "useless successful subschemas (middle)": {
+        "anyOf": Validator(
+            {
+                "anyOf": [
+                    *[True] * (NUM_USELESS // 2),
+                    subschema,
+                    *[True] * (NUM_USELESS // 2),
+                ],
+            },
+        ),
+        "allOf": Validator(
+            {
+                "allOf": [
+                    *[True] * (NUM_USELESS // 2),
+                    subschema,
+                    *[True] * (NUM_USELESS // 2),
+                ],
+            },
+        ),
+    },
+    "useless successful subschemas (end)": {
+        "anyOf": Validator({"anyOf": [*[True] * NUM_USELESS, subschema]}),
+        "allOf": Validator({"allOf": [*[True] * NUM_USELESS, subschema]}),
+    },
+    "useless failing subschemas (beginning)": {
+        "anyOf": Validator({"anyOf": [subschema, *[False] * NUM_USELESS]}),
+        "oneOf": Validator({"oneOf": [subschema, *[False] * NUM_USELESS]}),
+    },
+    "useless failing subschemas (middle)": {
+        "anyOf": Validator(
+            {
+                "anyOf": [
+                    *[False] * (NUM_USELESS // 2),
+                    subschema,
+                    *[False] * (NUM_USELESS // 2),
+                ],
+            },
+        ),
+        "oneOf": Validator(
+            {
+                "oneOf": [
+                    *[False] * (NUM_USELESS // 2),
+                    subschema,
+                    *[False] * (NUM_USELESS // 2),
+                ],
+            },
+        ),
+    },
+    "useless failing subschemas (end)": {
+        "anyOf": Validator({"anyOf": [*[False] * NUM_USELESS, subschema]}),
+        "oneOf": Validator({"oneOf": [*[False] * NUM_USELESS, subschema]}),
+    },
+}
+
+if __name__ == "__main__":
+    runner = Runner()
+
+    runner.bench_func("baseline valid", lambda: baseline.is_valid(valid))
+    runner.bench_func("baseline invalid", lambda: baseline.is_valid(invalid))
+
+    for group, applicators in by_name.items():
+        for applicator, validator in applicators.items():
+            runner.bench_func(
+                f"{group}: {applicator} valid",
+                lambda validator=validator: validator.is_valid(valid),
+            )
+            runner.bench_func(
+                f"{group}: {applicator} invalid",
+                lambda validator=validator: validator.is_valid(invalid),
+            )