better error messages

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "polars-h3"
-version = "0.4.3"  
+version = "0.4.4"  
 edition = "2021"
 
 [lib]
@@ -16,6 +16,7 @@ h3o = "0.4.0"
 rayon = "1.10.0"
 
 [profile.release]
+panic = "unwind"
 lto = true
 
 

benchmarks/engine.py

@@ -7,10 +7,11 @@
 - Attempted to also benchmark H3-Pandas, but project appears to be abandoned and doesn't work with h3 >= 4.0.0.
 """
 
-import json
 import random
+import statistics
 import time
-from dataclasses import asdict, dataclass, field
+from collections import defaultdict
+from dataclasses import dataclass, field
 from typing import Literal
 
 import duckdb
@@ -33,8 +34,10 @@
 class BenchmarkResult:
     library: Library
     name: str
-    seconds: float
+    avg_seconds: float
     num_rows: int
+    num_iterations: int
+    std_seconds: float = 0.0
 
     @property
     def num_rows_human(self) -> str:
@@ -46,9 +49,10 @@ def num_rows_human(self) -> str:
             return f"{self.num_rows / 1_000_000:,.0f}M"
 
     def __repr__(self) -> str:
-        return (
-            f"{self.library}::{self.name}::{self.num_rows_human} = {self.seconds:.2f}s"
-        )
+        if self.num_iterations == 0:
+            return f"{self.library}::{self.name}::{self.num_rows_human} = {self.avg_seconds:.2f}s"
+        else:
+            return f"{self.library}::{self.name}::{self.num_rows_human} = {self.avg_seconds:.2f}s ± {self.std_seconds:.2f}s"
 
 
 @dataclass
@@ -276,10 +280,11 @@ def run_all(
             for library in libraries:
                 func = config["funcs"][library]
 
-                start = time.perf_counter()
+                perf_times = []
                 for _ in range(self.config.num_iterations):
+                    start = time.perf_counter()
                     result_df = func(df.head(num_rows))
-                end = time.perf_counter()
+                    perf_times.append(time.perf_counter() - start)
 
                 if self.config.verbose:
                     print(f"Library: {library}")
@@ -290,8 +295,12 @@ def run_all(
                     BenchmarkResult(
                         name=func_name,
                         library=library,  # type: ignore
-                        seconds=(end - start),
+                        avg_seconds=statistics.mean(perf_times),
+                        std_seconds=statistics.stdev(perf_times)
+                        if len(perf_times) > 1
+                        else 0,
                         num_rows=num_rows,
+                        num_iterations=self.config.num_iterations,
                     )
                 )
             print("done...")
@@ -644,30 +653,61 @@ def _get_grid_paths_py_h3(self, df: pl.DataFrame) -> pl.DataFrame:
         )
 
 
+def _pretty_print_avg_results(results: list[BenchmarkResult]):
+    by_name = defaultdict(list)
+
+    for d in results:
+        by_name[d.name].append(d)
+
+    multiples = []
+    for speeds in by_name.values():
+        fastest = min(v.avg_seconds for v in speeds)
+        for v in speeds:
+            multiples.append((v.library, v.avg_seconds / fastest))
+
+    by_lib = defaultdict(list)
+    for lib, mult in multiples:
+        by_lib[lib].append(mult)
+
+    median_by_lib = {lib: round(statistics.median(ms), 2) for lib, ms in by_lib.items()}
+    avg_by_lib = {lib: round(sum(ms) / len(ms), 2) for lib, ms in by_lib.items()}
+
+    print("\n\n======= Benchmark Final Results =======\n")
+    print(f"{'Library':<10} {'Median':<8} {'Average':<8}")
+    print("-" * 26)
+    for lib in median_by_lib:
+        print(f"{lib:<10} {median_by_lib[lib]:<8} {avg_by_lib[lib]:<8}")
+
+
 if __name__ == "__main__":
     fast_factor = 1
     param_config = ParamConfig(
         resolution=9,
         grid_ring_distance=3,
-        num_iterations=1,
+        num_iterations=3,
         libraries="all",
         difficulty_to_num_rows={
             "basic": 10_000_000 // fast_factor,
             "medium": 10_000_000 // fast_factor,
             "complex": 100_000 // fast_factor,
         },
-        functions=["latlng_to_cell"],
+        # functions=["latlng_to_cell"],
         # verbose=True,
     )
     benchmark = Benchmark(config=param_config)
     results = benchmark.run_all()
     prev_func = None
     for result in results:
         if prev_func != result.name:
-            print(f"\n{result.name}")
+            print(f"\n{result.name} (num_iterations={param_config.num_iterations})")
             prev_func = result.name
         print(result)
 
+    _pretty_print_avg_results(results)
+
+    import json
+    from dataclasses import asdict
+
     if param_config.functions == "all":
         with open("benchmarks/benchmarks-results.json", "w") as f:
             json.dump([asdict(r) for r in results], f, indent=2)

polars_h3/__init__.py

@@ -41,6 +41,7 @@
     cell_area,
     edge_length,
     get_num_cells,
+    get_pentagons,
     great_circle_distance,
 )
 from .core.traversal import (
@@ -97,6 +98,7 @@
     "directed_edge_to_boundary",
     "great_circle_distance",
     "average_hexagon_area",
+    "get_pentagons",
     "cell_area",
     "edge_length",
     "average_hexagon_edge_length",