Optimize DateToString converter (#40)

* Optimize DateToString converter

* Configurable converters (special flag for unsafe converters)

* More comprehensive testing for unsafe DateToString converter

.gitignore

@@ -3,3 +3,5 @@ fq-connector-go-tests
 coverage*
 scripts/bench/postgresql
 __pycache__
+*.png
+*.test

app/bench/analyze.py

@@ -1,8 +1,9 @@
 #!/usr/bin/env python3
-
+from typing import Dict
 from pathlib import Path
 import json
 
+import matplotlib
 import matplotlib.pyplot as plt
 import pandas as pd
 
@@ -15,39 +16,66 @@ def read_report(path: Path) -> pd.Series:
             "bytes_internal_rate": data["bytes_internal_rate"],
             "bytes_arrow_rate": data["bytes_arrow_rate"],
             "rows_rate": data["rows_rate"],
+            "cpu_utilization": data["cpu_utilization"],
         }
     )
 
 
-def make_dataframe(result_dir: Path) -> pd.DataFrame:
-    series = [read_report(path) for path in result_dir.glob("*.json")]
-    return pd.DataFrame(series)
+def make_dataframe(result_dirs: Dict[str, Path]) -> pd.DataFrame:
+    dfs = []
+    for key, result_dir in result_dirs.items():
+        series = [read_report(path) for path in result_dir.glob("*.json")]
+        df = pd.DataFrame(series).sort_values("columns")
+        df["key"] = key
+        dfs.append(df)
+
+    return pd.concat(dfs)
+
 
+def draw_subplot(
+    df_: pd.DataFrame, label: str, y_column: str, ax: matplotlib.figure.Figure
+) -> matplotlib.figure.Figure:
+    ax.set_ylabel(label)
+    ax.set_xlabel("Number of columns to SELECT")
 
-def draw_plot(df: pd.DataFrame, result_dir: Path) -> pd.Series:
-    fig, ax1 = plt.subplots()
+    keys = {
+        "baseline": "red",
+        "optimized": "blue",
+    }
 
-    ax1.set_xlabel("Number of columns in SELECT")
-    ax1.set_ylabel("Throughput, MB/sec", color="red")
-    ax1.scatter(df["columns"], df["bytes_internal_rate"], color="red")
-    ax1.tick_params(axis="y", labelcolor="red")
+    for key, color in keys.items():
+        df = df_.loc[df_["key"] == key]
+        ax.plot(df["columns"], df[y_column], color=color, label=key)
 
-    ax2 = ax1.twinx()
+    return ax
 
-    ax2.set_ylabel("Throughput, rows/sec", color="blue")
-    ax2.scatter(df["columns"], df["rows_rate"], color="blue")
-    ax2.tick_params(axis="y", labelcolor="blue")
 
-    fig.savefig(result_dir.joinpath("report.png"))
+def draw_plot(df: pd.DataFrame) -> pd.Series:
+    fig, axes = plt.subplots(nrows=1, ncols=3, figsize=(12, 4))
+    fig.subplots_adjust(bottom=0.25, wspace=0.5)
+    draw_subplot(df, "Throughput, MB/sec", "bytes_internal_rate", axes[0])
+    draw_subplot(df, "Throughput, rows/sec", "rows_rate", axes[1])
+    ax = draw_subplot(df, "CPU Utilization, %", "cpu_utilization", axes[2])
+
+    handles, labels = ax.get_legend_handles_labels()
+    fig.legend(handles, labels, loc="lower right")
+    fig.suptitle("Reading TPC-H S-10 Lineitem from PostgreSQL", fontsize=14)
+
+    fig.savefig("report.png")
 
 
 def main():
-    result_dir = Path(
-        "/home/vitalyisaev/projects/fq-connector-go/scripts/bench/postgresql/results/columns/"
-    )
-    df = make_dataframe(result_dir)
+    result_dirs = {
+        "baseline": Path(
+            "/home/vitalyisaev/projects/fq-connector-go/scripts/bench/postgresql/results/columns_baseline/"
+        ),
+        "optimized": Path(
+            "/home/vitalyisaev/projects/fq-connector-go/scripts/bench/postgresql/results/columns/"
+        ),
+    }
+    df = make_dataframe(result_dirs)
     print(df)
-    draw_plot(df, result_dir)
+    draw_plot(df)
 
 
 if __name__ == "__main__":

app/bench/test_case_runner.go

@@ -36,6 +36,11 @@ func newTestCaseRunner(
 		server.WithPprofServerConfig(&config.TPprofServerConfig{
 			Endpoint: &api_common.TEndpoint{Host: "localhost", Port: 50052},
 		}),
+		server.WithConversionConfig(
+			&config.TConversionConfig{
+				UseUnsafeConverters: true,
+			},
+		),
 	)
 
 	if err != nil {