Merge pull request #5726 from Belagum/px-unsigned-int-continuous

Fix px treating unsigned integer columns as categorical

Author: emilykl

CHANGELOG.md

@@ -9,6 +9,7 @@ This project adheres to [Semantic Versioning](http://semver.org/).
 - Add support for custom tick values in `mpl_to_plotly` when the matplotlib tick positions don't follow an arithmetic progression, including custom tick labels and tick values on date axes [[#5262](https://github.com/plotly/plotly.py/pull/5262)], with thanks to @robertoffmoura for the contribution!
 
 ### Fixed
+- Fix Plotly Express treating unsigned integer columns (`uint8`, `uint16`, `uint32`, `uint64`) in pandas DataFrames as categorical, which caused `px.bar` and other functions to pick the wrong orientation and render unexpected plots [[#4291](https://github.com/plotly/plotly.py/issues/4291), [#4344](https://github.com/plotly/plotly.py/issues/4344)], with thanks to @Belagum for the contribution!
 - Fix `mpl_to_plotly` not setting `paper_bgcolor` and `plot_bgcolor` from the matplotlib figure and axes backgrounds, so converted figures match the source figure's background colors [[#5285](https://github.com/plotly/plotly.py/pull/5285)], with thanks to @robertoffmoura for the contribution!
 - Fix rendering issue causing a too-large div when calling `Figure.show()` in Google Colab [[#5718](https://github.com/plotly/plotly.py/pull/5718)]
 - Fix the sphinx-gallery scraper so that it generates thumbnails for figures shown with `fig.show()` or displayed as the last expression of a code block [[#5701](https://github.com/plotly/plotly.py/pull/5701)], with thanks to @larsoner for the contribution!

plotly/express/_core.py

@@ -149,7 +149,7 @@ def _is_continuous(df: nw.DataFrame, col_name: str) -> bool:
         # fastpath for pandas: Narwhals' Series.dtype has a bit of overhead, as it
         # tries to distinguish between true "object" columns, and "string" columns
         # disguised as "object". But here, we deal with neither.
-        return df_native[col_name].dtype.kind in "ifc"
+        return df_native[col_name].dtype.kind in "iufc"
     return df.get_column(col_name).dtype.is_numeric()
 
 

tests/test_optional/test_px/test_px_input.py

@@ -7,7 +7,7 @@
 import pytest
 from packaging import version
 import unittest.mock as mock
-from plotly.express._core import build_dataframe
+from plotly.express._core import build_dataframe, _is_continuous
 from plotly import optional_imports
 from pandas.testing import assert_frame_equal
 import sys
@@ -624,6 +624,19 @@ def test_auto_orient_x_and_y(fn, x, y, result):
             assert fn(x=series[x], y=series[y]).data[0].orientation == result
 
 
+@pytest.mark.parametrize("dtype", [nw.UInt8, nw.UInt16, nw.UInt32, nw.UInt64])
+def test_auto_orient_unsigned_int(constructor, dtype):
+    df = nw.from_native(constructor(dict(category=["a", "b", "c"], value=[1, 2, 3])))
+    df = df.with_columns(nw.col("value").cast(dtype)).to_native()
+
+    assert _is_continuous(nw.from_native(df, eager_only=True), "value")
+    fig = px.bar(df, x="category", y="value")
+    assert fig.data[0].orientation == "v"
+    assert list(fig.data[0].y) == [1, 2, 3]
+    fig = px.bar(df, x="value", y="category")
+    assert fig.data[0].orientation == "h"
+
+
 def test_histogram_auto_orient():
     numerical = [1, 2, 3, 4]
     assert px.histogram(x=numerical, nbins=5).data[0].nbinsx == 5