Merge pull request #16 from gerrymanoim/benchmark-list-access

BENCH: Add random Array indexing benchmark

.github/workflows/benchmark.yml

@@ -38,7 +38,7 @@ jobs:
         PYTHONWARNINGS: ignore:DEPRECATION::pip._internal.cli.base_command
       run: |
         python -m pip install --upgrade pip
-        pip install numpy libpy
+        pip install numpy libpy pybind11
     - name: Install
       if: steps.check.outputs.triggered == 'true'
       run: |

libpy_simdjson/tests/test_benchmark.py

@@ -9,6 +9,7 @@
 from ujson import loads as ujson_loads
 from rapidjson import loads as rapidjson_loads
 from simdjson import loads as pysimdjson_loads
+from simdjson import parse as pysimdjson_parse
 from libpy_simdjson import loads as libpy_simdjson_loads
 
 from libpy_simdjson import Array
@@ -29,7 +30,8 @@ def libpy_simdjson_as_py_obj(input):
 @pytest.mark.parametrize(
     ["group", "func"],
     [
-        ("pysimdjson", pysimdjson_loads),
+        ("pysimdjson_as_py_obj", pysimdjson_loads),
+        ("pysimdjson", pysimdjson_parse),
         ("orjson", orjson_loads),
         ("ujson", ujson_loads),
         ("rapidjson", rapidjson_loads),
@@ -91,3 +93,39 @@ def py_test_func(doc):
         content = f.read()
         doc = read_func(content)
         benchmark(bench_func, doc)
+
+
+@pytest.mark.parametrize(
+    ["group", "read_func"],
+    [
+        ("python_json", json_loads),
+        ("libpy_simdjson", libpy_simdjson_loads),
+    ],
+)
+def test_benchmark_list_access(group, read_func, benchmark):
+    benchmark.group = "Random list access"
+    benchmark.extra_info["group"] = group
+
+    random.seed(999)
+
+    def simd_test_func(doc):
+        selection = random.randrange(10_000)
+        doc[selection]
+
+    def py_test_func(doc):
+        selection = random.randrange(10_000)
+        doc[selection]
+
+    if group == "libpy_simdjson":
+        bench_func = simd_test_func
+    elif group == "python_json":
+        bench_func = py_test_func
+    else:
+        raise ValueError("unknown group for direct access test")
+
+    json_doc = JSON_FIXTURES_DIR / "numbers.json"
+    with json_doc.open('rb') as f:
+        content = f.read()
+        doc = read_func(content)
+        benchmark(bench_func, doc)
+

setup.py

@@ -80,7 +80,6 @@ def extension(*args, **kwargs):
             "orjson",
             "python-rapidjson",
             "pysimdjson",
-            "pybind11", # used by pysimdjson if wheels do not exist
             "ujson",
         ],
     },