New ome-arrow performance benchmarks

CITATION.cff

@@ -17,7 +17,7 @@ authors:
   - given-names: Gregory
     family-names: Way
     orcid: 'https://orcid.org/0000-0002-0503-9348'
-repository-code: "https://github.com/wayscience/ome-arrow"
+repository-code: "https://github.com/wayscience/ome-arrow-benchmarks"
 abstract: >-
   Benchmarking OME Arrow through Parquet, Vortex, LanceDB, and more.
 keywords:

pyproject.toml

@@ -23,7 +23,8 @@ dependencies = [
   "matplotlib>=3.10.9",
   "numcodecs>=0.16.5",
   "numpy>=2.4.6",
-  "ome-arrow>=0.0.9",
+  "ome-arrow>=0.0.10",
+  "ome-iris>=0.0.4",
   "ome-zarr>=0.16",
   "pandas>=3.0.3",
   "poethepoet>=0.46",
@@ -52,6 +53,7 @@ paths = [ "src/benchmarks" ]
 tasks.lab.shell = "uv venv && uv sync && uv run jupyter lab"
 tasks.run-benchmarks.shell = """\
   uv run python src/benchmarks/compare_parquet_vortex_lance.py && uv run python \
-  src/benchmarks/compare_parquet_vortex_lance_ome.py && uv run python src/benchmarks/compare_ome_arrow_only.py\
+  src/benchmarks/compare_parquet_vortex_lance_ome.py && uv run python src/benchmarks/compare_ome_arrow_only.py && uv \
+  run python src/benchmarks/compare_ome_iris_joins.py\
   """
 tasks.setup.shell = "uv venv && uv sync"

src/benchmarks/compare_ome_arrow_only.py

@@ -1,4 +1,4 @@
-"""OME-Arrow single-column benchmarks
+"""OME image-column storage benchmarks
 
 Compare on-disk formats for a table containing only `row_id` and one OME-Arrow
 image column, plus a separate benchmark for a directory-per-image OME-Zarr
@@ -33,11 +33,11 @@
 from ome_arrow import OMEArrow
 
 pd.set_option("display.precision", 4)
-PLOT_TITLE = f"{(__doc__ or 'OME-Arrow single-column benchmarks').splitlines()[0]} (lower is better)"
+PLOT_TITLE = "One OME-Arrow Image Column: Table Backends vs Directory Baselines\n(lower is better)"
 DATA_DIR = Path("data")
 DATA_DIR.mkdir(exist_ok=True)
-IMAGES_DIR = Path("images")
-IMAGES_DIR.mkdir(exist_ok=True)
+FIGURES_DIR = Path("figures")
+FIGURES_DIR.mkdir(exist_ok=True)
 SUMMARY_PARQUET = DATA_DIR / "compare_ome_arrow_only_summary.parquet"
 RUNS_PARQUET = DATA_DIR / "compare_ome_arrow_only_runs.parquet"
 RUN_BENCHMARKS = not (SUMMARY_PARQUET.exists() and RUNS_PARQUET.exists())
@@ -623,6 +623,15 @@ def tiff_random_read(indices, base_path=TIFF_DIR):
     "OME-Zarr (dir-per-image)": "#7A5A3C",
 }
 colors = [COLOR_MAP.get(name, "#BAB0AC") for name in summary["format"]]
+DISPLAY_LABELS = {
+    "Parquet (pyarrow, zstd)": "PQ\npyarrow",
+    "Parquet (duckdb, zstd)": "PQ\nduckdb",
+    "Lance (lancedb)": "Lance",
+    "Vortex": "Vortex",
+    "DuckDB (file table)": "DuckDB",
+    "TIFF (dir-per-image)": "TIFF",
+    "OME-Zarr (dir-per-image)": "OME-Zarr",
+}
 
 
 def label_bars(ax, bars, fmt="%.3f"):
@@ -662,13 +671,13 @@ def label_bars(ax, bars, fmt="%.3f"):
     bars = ax.bar(x, summary[col], color=colors)
     ax.set_title(title)
     ax.set_xticks(x)
-    ax.set_xticklabels(summary["format"], rotation=30, ha="right")
+    ax.set_xticklabels([DISPLAY_LABELS.get(name, name) for name in summary["format"]])
     ax.grid(axis="y", linestyle=":", alpha=0.5)
     if col != "size_mb":
         ax.set_ylabel("seconds")
     else:
         ax.set_ylabel("MB")
     label_bars(ax, bars)
 
-fig.savefig(IMAGES_DIR / "compare_ome_arrow_only_summary.png", dpi=150)
+fig.savefig(FIGURES_DIR / "compare_ome_arrow_only_summary.png", dpi=150)
 plt.close(fig)