fix(python): Consider the original dtypes when selecting columns in write_excel function

py-polars/polars/dataframe/frame.py

@@ -3473,6 +3473,12 @@ def write_excel(
         wb, ws, can_close = _xl_setup_workbook(workbook, worksheet)
         df, is_empty = self, not len(self)
 
+        # The _xl_setup_table_columns function in the below section
+        # converts all collection types (e.g. List, Struct, Object) to strings
+        # Hence, we need to store the original schema so that it can be used
+        # when selecting columns using column selectors based on datatypes
+        df_original = df.clear()
+
         # setup table format/columns
         fmt_cache = _XLFormatCache(wb)
         column_formats = column_formats or {}
@@ -3540,7 +3546,7 @@ def write_excel(
         elif isinstance(hidden_columns, str):
             hidden = {hidden_columns}
         else:
-            hidden = set(_expand_selectors(df, hidden_columns))
+            hidden = set(_expand_selectors(df_original, hidden_columns))
 
         # Autofit section needs to be present above column_widths section
         # to ensure that parameters provided in the column_widths section
@@ -3558,7 +3564,7 @@ def write_excel(
             column_widths = dict.fromkeys(df.columns, column_widths)
         else:
             column_widths = _expand_selector_dicts(  # type: ignore[assignment]
-                df, column_widths, expand_keys=True, expand_values=False
+                df_original, column_widths, expand_keys=True, expand_values=False
             )
         column_widths = _unpack_multi_column_dict(column_widths or {})  # type: ignore[assignment]
 

py-polars/tests/unit/io/test_spreadsheet.py

@@ -1350,3 +1350,53 @@ def test_drop_empty_rows(
         drop_empty_rows=False,
     )
     assert df3.shape == (10, 4)
+
+
+def test_write_excel_select_col_dtype() -> None:
+    from openpyxl import load_workbook
+    from xlsxwriter import Workbook
+
+    def get_col_widths(wb_bytes: BytesIO) -> dict[str, int]:
+        return {
+            k: round(v.width)
+            for k, v in load_workbook(wb_bytes).active.column_dimensions.items()
+        }
+
+    df = pl.DataFrame(
+        {
+            "name": [["Alice", "Ben"], ["Charlie", "Delta"]],
+            "col2": ["Hi", "Bye"],
+        }
+    )
+
+    # column_widths test:
+    # pl.List(pl.String)) datatype should not match column with no list
+    check = BytesIO()
+    with Workbook(check) as wb:
+        df.write_excel(wb, column_widths={cs.by_dtype(pl.List(pl.String)): 300})
+
+    assert get_col_widths(check) == {"A": 43}
+
+    # column_widths test:
+    # pl.String datatype should not match column with list
+    check = BytesIO()
+    with Workbook(check) as wb:
+        df.write_excel(wb, column_widths={cs.by_dtype(pl.String): 300})
+
+    assert get_col_widths(check) == {"B": 43}
+
+    # hidden_columns test:
+    # pl.List(pl.String)) datatype should not match column with no list
+    check = BytesIO()
+    with Workbook(check) as wb:
+        df.write_excel(wb, hidden_columns=cs.by_dtype(pl.List(pl.String)))
+
+    assert get_col_widths(check) == {"A": 0}
+
+    # hidden_columns test:
+    # pl.String datatype should not match column with list
+    check = BytesIO()
+    with Workbook(check) as wb:
+        df.write_excel(wb, hidden_columns=cs.by_dtype(pl.String))
+
+    assert get_col_widths(check) == {"B": 0}