Repository navigation
fix(bar): percentage_by_bar_group fails when an x value repeats #598
Description
What: bar.plot(..., data_label_format="percentage_by_bar_group") fails when an x value repeats: ValueError: zip() argument 2 is shorter than argument 1. It fails the same way with sort_order set, and when the repeated values are the index. Checked on main @ c5f4b2a.
A repeated x is valid input: pandas draws one bar per row, and the other label formats work. The docstring says this mode shows "each bar's share of its x-axis group".
Cause: bar.py:219 computes df.groupby(x_col)[value_col].sum().sum(axis=1), one total per x value. _generate_percentage_labels then zips that against every bar with strict=True. The fix needs one denominator per row, e.g. groupby(x_col)[value_col].transform("sum").sum(axis=1).
Failing test:
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
import pandas as pd
import pytest
from openretailscience.plots import bar
@pytest.fixture(autouse=True)
def cleanup_figures():
yield
plt.close("all")
class TestBarPercentageByBarGroupRepeatedX:
@pytest.mark.parametrize("use_index", [False, True])
def test_labels_are_share_of_x_group(self, use_index):
"""Leeds 10/30 splits 25%/75%; York 20/20 splits 50%/50%."""
df = pd.DataFrame(
{
"store": ["Leeds", "Leeds", "York", "York"],
"channel": ["web", "shop", "web", "shop"],
"revenue": [10.0, 30.0, 20.0, 20.0],
},
)
x_col = "store"
if use_index:
df = df.set_index("store")
x_col = None
ax = bar.plot(df, value_col="revenue", x_col=x_col, data_label_format="percentage_by_bar_group")
labels = [float(t.get_text().rstrip("%")) for t in ax.texts]
assert labels == [25.0, 75.0, 50.0, 50.0]
Activity