Merge pull request #393 from OpenCOMPES/bugfix-numba-binning

Exception handling for binning when df cols are of non-numeric types

sed/binning/binning.py

@@ -170,7 +170,14 @@ def bin_partition(
                 apply_jitter_on_column(sel_part, amp * binsize, col, mode)
         vals = sel_part.values
     else:
-        vals = part.values[:, col_id]
+        vals = part.iloc[:, col_id].values
+    if vals.dtype == "object":
+        raise ValueError(
+            "Binning requires all binned dataframe columns to be of numeric type. "
+            "Encountered data types were "
+            f"{[part.columns[id] + ': ' + str(part.iloc[:, id].dtype) for id in col_id]}. "
+            "Please make sure all axes data are of numeric type.",
+        )
     if hist_mode == "numba":
         hist_partition, edges = numba_histogramdd(
             vals,

tests/test_binning.py

@@ -484,6 +484,20 @@ def test_bin_partition() -> None:
     assert np.allclose(cast(np.ndarray, res), res1)
 
 
+def test_non_numeric_dtype_error() -> None:
+    """Test bin_partition function"""
+    pdf = sample_pdf.astype({"x": "string", "y": "int32", "z": "int32"})
+    with pytest.raises(ValueError) as err:
+        _ = bin_partition(
+            part=pdf,
+            bins=bins,  # type: ignore[arg-type]
+            axes=columns,
+            ranges=ranges,
+            skip_test=False,
+        )
+    assert "Encountered data types were ['x: string', 'y: int32', 'z: int32']" in str(err.value)
+
+
 def test_bin_dataframe() -> None:
     """Test bin_dataframe function"""
     res = bin_dataframe(df=sample_ddf, bins=bins, axes=columns, ranges=ranges)