BUG: to_excel inserts invalid data if a level in a MultiIndex is None

[musshorn]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import pandas as pd
import numpy as np

Cols = pd.MultiIndex.from_tuples([("Item", "Type 1", "Class A"), ("Item", "Type 1", "Class B"), ("Item", "Type 2", "Class A"), ("Item", "Type 2", "Class B"), ("Item", "Type 3", None)])

df = pd.DataFrame(np.random.randn(10, 5), columns=Cols)
print(df)
df.to_excel("Test.xlsx")

Issue Description

When a level in a MultiIndex is None, the resulting dataframe when exported to excel has an invalid level inserted replacing the None. See Image.

Expected Behavior

I would expect the result from to_excel should reflect the levels in the MultiIndex correctly.

Installed Versions

pd.show_versions()

INSTALLED VERSIONS

commit : 2e218d1
python : 3.10.0.final.0
python-bits : 64
OS : Windows
OS-release : 10
Version : 10.0.19044
machine : AMD64
processor : Intel64 Family 6 Model 79 Stepping 1, GenuineIntel
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : English_Australia.1252

pandas : 1.5.3
numpy : 1.22.4
pytz : 2022.2.1
dateutil : 2.8.2
setuptools : 57.4.0
pip : 23.0
Cython : 0.29.32
pytest : 7.2.0
hypothesis : None
sphinx : 4.5.0
blosc : None
feather : None
xlsxwriter : 3.0.1
lxml.etree : 4.9.1
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 3.1.2
IPython : 8.1.1
pandas_datareader: None
bs4 : 4.10.0
bottleneck : None
brotli : 1.0.9
fastparquet : None
fsspec : None
gcsfs : None
matplotlib : 3.5.1
numba : None
numexpr : 2.8.4
odfpy : None
openpyxl : 3.1.0
pandas_gbq : None
pyarrow : None
pyreadstat : None
pyxlsb : None
s3fs : None
scipy : 1.8.0
snappy : None
sqlalchemy : None
tables : 3.7.0
tabulate : 0.8.9
xarray : None
xlrd : None
xlwt : 1.3.0
zstandard : 0.17.0
tzdata : None

[phofl]

I guess this is related to forward filling for round tripping purposes (see read_excel index_col documentation). But investigations welcome

[musshorn]

I think

pandas/pandas/io/formats/excel.py

Lines 641 to 660 in 13db83a

	for lnum, (spans, levels, level_codes) in enumerate(
	zip(level_lengths, columns.levels, columns.codes)
	):
	values = levels.take(level_codes)
	for i, span_val in spans.items():
	mergestart, mergeend = None, None
	if span_val > 1:
	mergestart, mergeend = lnum, coloffset + i + span_val
	yield CssExcelCell(
	row=lnum,
	col=coloffset + i + 1,
	val=values[i],
	style=self.header_style,
	css_styles=getattr(self.styler, "ctx_columns", None),
	css_row=lnum,
	css_col=i,
	css_converter=self.style_converter,
	mergestart=mergestart,
	mergeend=mergeend,
	)

is the offending loop. When it reaches the last level, levels.take(level_codes) gets passed [0, 1, 0, 1, -1] which sets values to ['Class A', 'Class B', 'Class A', 'Class B', 'Class B'] and that's ultimately the row that gets written to Excel.

[ghost]

pandas/pandas/core/indexes/base.py

Line 1052 in b9a4335

allow_fill = self._maybe_disallow_fill(allow_fill, fill_value, indices)

This variable is set to False for your case. When I manually set it to True in my debugger I get the expected Excel output.

Now it's left to figure out why it's set to False automatically...

Can I take it?

[musshorn]

Yeah all yours, I was just trying to provide some starting research for whoever took it up

[ghost]

take