Make resamply.py in ASV benchmarks dir

python/benchmarks/resample.py

@@ -0,0 +1,121 @@
+"""
+Copyright 2024 Man Group Operations Limited
+
+Use of this software is governed by the Business Source License 1.1 included in the file licenses/BSL.txt.
+
+As of the Change Date specified in that file, in accordance with the Business Source License, use of this software will be governed by the Apache License, version 2.0.
+"""
+from arcticdb import Arctic
+from arcticdb.version_store.processing import QueryBuilder
+
+from .common import *
+
+
+class LocalQueryBuilderFunctions:
+    number = 5
+    timeout = 6000
+    LIB_NAME = "query_builder"
+    CONNECTION_STRING = "lmdb://query_builder?map_size=5GB"
+
+    params = [1_000_000, 10_000_000]
+    param_names = ["num_rows"]
+
+    def setup_cache(self):
+        self.ac = Arctic(LocalQueryBuilderFunctions.CONNECTION_STRING)
+
+        num_rows = LocalQueryBuilderFunctions.params
+        self.lib_name = LocalQueryBuilderFunctions.LIB_NAME
+        self.ac.delete_library(self.lib_name)
+        lib = self.ac.create_library(self.lib_name)
+        for rows in num_rows:
+            lib.write(f"{rows}_rows", generate_benchmark_df(rows))
+
+    def teardown(self, num_rows):
+        del self.lib
+        del self.ac
+
+    def setup(self, num_rows):
+        self.ac = Arctic(LocalQueryBuilderFunctions.CONNECTION_STRING)
+        self.lib = self.ac[LocalQueryBuilderFunctions.LIB_NAME]
+
+    # Omit string columns in filtering/projection benchmarks to avoid time/memory being dominated by Python string
+    # allocation
+    def time_filtering_numeric(self, num_rows):
+        q = QueryBuilder()
+        # v3 is random floats between 0 and 100
+        q = q[q["v3"] < 1.0]
+        self.lib.read(f"{num_rows}_rows", columns=["v3"], query_builder=q)
+
+    def peakmem_filtering_numeric(self, num_rows):
+        q = QueryBuilder()
+        # v3 is random floats between 0 and 100
+        q = q[q["v3"] < 10.0]
+        self.lib.read(f"{num_rows}_rows", columns=["v3"], query_builder=q)
+
+    def time_filtering_string_isin(self, num_rows):
+        # Selects about 1% of the rows
+        k = num_rows // 1000
+        string_set = [f"id{str(i).zfill(3)}" for i in range(1, k + 1)]
+        q = QueryBuilder()
+        q = q[q["id1"].isin(string_set)]
+        self.lib.read(f"{num_rows}_rows", columns=["v3"], query_builder=q)
+
+    def peakmem_filtering_string_isin(self, num_rows):
+        # Selects about 1% of the rows
+        k = num_rows // 1000
+        string_set = [f"id{str(i).zfill(3)}" for i in range(1, k + 1)]
+        q = QueryBuilder()
+        q = q[q["id1"].isin(string_set)]
+        self.lib.read(f"{num_rows}_rows", columns=["v3"], query_builder=q)
+
+    def time_projection(self, num_rows):
+        q = QueryBuilder()
+        q = q.apply("new_col", q["v2"] * q["v3"])
+        self.lib.read(f"{num_rows}_rows", columns=["new_col"], query_builder=q)
+
+    def peakmem_projection(self, num_rows):
+        q = QueryBuilder()
+        q = q.apply("new_col", q["v2"] * q["v3"])
+        self.lib.read(f"{num_rows}_rows", columns=["new_col"], query_builder=q)
+
+    # The names are based on the queries used here: https://duckdblabs.github.io/db-benchmark/
+    # Don't rename to distinguish from other query tests as renaming makes it a new benchmark, losing historic results
+    def time_query_1(self, num_rows):
+        q = QueryBuilder()
+        q = q.groupby("id1").agg({"v1": "sum"})
+        self.lib.read(f"{num_rows}_rows", query_builder=q)
+
+    def peakmem_query_1(self, num_rows):
+        q = QueryBuilder()
+        q = q.groupby("id1").agg({"v1": "sum"})
+        self.lib.read(f"{num_rows}_rows", query_builder=q)
+
+    def time_query_3(self, num_rows):
+        q = QueryBuilder()
+        q = q.groupby("id3").agg({"v1": "sum", "v3": "sum"})
+        self.lib.read(f"{num_rows}_rows", query_builder=q)
+
+    def peakmem_query_3(self, num_rows):
+        q = QueryBuilder()
+        q = q.groupby("id3").agg({"v1": "sum", "v3": "sum"})
+        self.lib.read(f"{num_rows}_rows", query_builder=q)
+
+    def time_query_4(self, num_rows):
+        q = QueryBuilder()
+        q = q.groupby("id6").agg({"v1": "sum", "v2": "sum"})
+        self.lib.read(f"{num_rows}_rows", query_builder=q)
+
+    def peakmem_query_4(self, num_rows):
+        q = QueryBuilder()
+        q = q.groupby("id6").agg({"v1": "sum", "v2": "sum"})
+        self.lib.read(f"{num_rows}_rows", query_builder=q)
+
+    def time_query_adv_query_2(self, num_rows):
+        q = QueryBuilder()
+        q = q.groupby("id3").agg({"v1": "max", "v2": "min"})
+        self.lib.read(f"{num_rows}_rows", query_builder=q)
+
+    def peakmem_query_adv_query_2(self, num_rows):
+        q = QueryBuilder()
+        q = q.groupby("id3").agg({"v1": "max", "v2": "min"})
+        self.lib.read(f"{num_rows}_rows", query_builder=q)