BUG: read_excel not forward filling index columns when no names are given (#47496)

Author: phofl

doc/source/whatsnew/v1.5.0.rst

@@ -935,6 +935,7 @@ I/O
 - Bug in :func:`read_parquet` when ``engine="fastparquet"`` where the file was not closed on error (:issue:`46555`)
 - :meth:`to_html` now excludes the ``border`` attribute from ``<table>`` elements when ``border`` keyword is set to ``False``.
 - Bug in :func:`read_sas` with certain types of compressed SAS7BDAT files (:issue:`35545`)
+- Bug in :func:`read_excel` not forward filling :class:`MultiIndex` when no names were given (:issue:`47487`)
 - Bug in :func:`read_sas` returned ``None`` rather than an empty DataFrame for SAS7BDAT files with zero rows (:issue:`18198`)
 - Bug in :class:`StataWriter` where value labels were always written with default encoding (:issue:`46750`)
 - Bug in :class:`StataWriterUTF8` where some valid characters were removed from variable names (:issue:`47276`)

pandas/io/excel/_base.py

@@ -788,9 +788,27 @@ def parse(
 
             # If there is a MultiIndex header and an index then there is also
             # a row containing just the index name(s)
-            has_index_names = (
-                is_list_header and not is_len_one_list_header and index_col is not None
-            )
+            has_index_names = False
+            if is_list_header and not is_len_one_list_header and index_col is not None:
+
+                index_col_list: Sequence[int]
+                if isinstance(index_col, int):
+                    index_col_list = [index_col]
+                else:
+                    assert isinstance(index_col, Sequence)
+                    index_col_list = index_col
+
+                # We have to handle mi without names. If any of the entries in the data
+                # columns are not empty, this is a regular row
+                assert isinstance(header, Sequence)
+                if len(header) < len(data):
+                    potential_index_names = data[len(header)]
+                    potential_data = [
+                        x
+                        for i, x in enumerate(potential_index_names)
+                        if not control_row[i] and i not in index_col_list
+                    ]
+                    has_index_names = all(x == "" or x is None for x in potential_data)
 
             if is_list_like(index_col):
                 # Forward fill values for MultiIndex index.

pandas/tests/io/data/excel/multiindex_no_index_names.xlsx

@@ -396,3 +396,17 @@ def test_ints_spelled_with_decimals(datapath, ext):
     result = pd.read_excel(path)
     expected = DataFrame(range(2, 12), columns=[1])
     tm.assert_frame_equal(result, expected)
+
+
+def test_read_multiindex_header_no_index_names(datapath, ext):
+    # GH#47487
+    path = datapath("io", "data", "excel", f"multiindex_no_index_names{ext}")
+    result = pd.read_excel(path, index_col=[0, 1, 2], header=[0, 1, 2])
+    expected = DataFrame(
+        [[np.nan, "x", "x", "x"], ["x", np.nan, np.nan, np.nan]],
+        columns=pd.MultiIndex.from_tuples(
+            [("X", "Y", "A1"), ("X", "Y", "A2"), ("XX", "YY", "B1"), ("XX", "YY", "B2")]
+        ),
+        index=pd.MultiIndex.from_tuples([("A", "AA", "AAA"), ("A", "BB", "BBB")]),
+    )
+    tm.assert_frame_equal(result, expected)