Some benchmarks with arrow strings

tests/benchmarks/test_arrow.py

@@ -0,0 +1,58 @@
+import pandas as pd
+import pytest
+
+from ..utils_test import cluster_memory, timeseries_of_size, wait
+
+
+@pytest.fixture(params=[True, False], ids=["string[pyarrow]", "object"])
+def series_with_client(request, small_client):
+    memory = cluster_memory(small_client)
+    df = timeseries_of_size(memory)
+    series = df.name
+    if request.param:
+        series = series.astype(pd.StringDtype("pyarrow"))
+    series = series.persist()
+    yield series, small_client
+
+
+def test_unique(series_with_client):
+    """Find unique values"""
+    series, client = series_with_client
+    result = series.unique()
+    wait(result, client, 10 * 60)
+
+
+def test_contains(series_with_client):
+    """String contains"""
+    series, client = series_with_client
+    result = series.str.contains("a")
+    wait(result, client, 10 * 60)
+
+
+def test_startswith(series_with_client):
+    """String starts with"""
+    series, client = series_with_client
+    result = series.str.startswith("B")
+    wait(result, client, 10 * 60)
+
+
+def test_upper(series_with_client):
+    """String upper"""
+    series, client = series_with_client
+    result = series.str.upper()
+    wait(result, client, 10 * 60)
+
+
+def test_filter(series_with_client):
+    """How fast can we filter the Series"""
+    series, client = series_with_client
+    name = series.head(1)[0]  # Get first name that appears
+    result = series[series == name]
+    wait(result, client, 10 * 60)
+
+
+def test_value_counts(series_with_client):
+    """Value counts on string values"""
+    series, client = series_with_client
+    result = series.value_counts()
+    wait(result, client, 10 * 60)