[BUG] conversion to nested_univ is unexpected

[weenerplasticsgroup]

Describe the bug
conversion to nested_univ is unexpected

To Reproduce
I start with this table

from sktime.datatypes import get_examples
X = get_examples(mtype="pd-multiindex", as_scitype="Panel")[0]
X_trafo = X.reset_index(level=0, drop=True).rename(columns=dict(map(reversed, enumerate(X.columns)))).stack().to_frame().groupby(level=[0,1]).first()
X_trafo

		0
timepoints
0	0	1
	1	4
1	0	2
	1	5
2	0	3
	1	6

Now I convert to nested_univ

from sktime.datatypes import convert_to
X_new = convert_to(X_trafo, "nested_univ")
X_new

	0
timepoints
0	0 1 1 4 Name: 0, dtype: int64
1	0 2 1 5 Name: 0, dtype: int64
2	0 3 1 6 Name: 0, dtype: int64

Expected behavior
This is unexpected, I have 4 values where I expect two

I expect
Time 0 series 1 4
Time 1 series 2 5
etc.

Additional context

This example is very similar but does not give the unexpected result

from sktime.datatypes import convert_to
from sktime.datasets import load_arrow_head
X, _ = load_arrow_head(return_X_y=True)
X_new = convert_to(X, to_type="pd-multiindex").swaplevel()
convert_to(X_new.sort_index(), to_type="nested_univ")

Versions
sktime 2.9

[fkiraly]

This is just how it prints out, there are actually only two values inside, in the pd.Series inside the cell.

It prints the index values, e.g., [0, 1] in the top left cell, next to the series values, e.g., [1, 4] in the top left cell.

PS, we do no longer recommend using the nested_univ specification, because pandas 2 no longer supports the commands that would render manipulation of the container easy, and it is not in a supported pandas API.

[hstarmans]

The format is different, it looks strange. I tested it with X.iloc[0].values and ended up with four, but will check again tomorrow

[fkiraly]

?? ok, let me run it. I was pretty sure it is just printing. Give me a sec.

[fkiraly]

Yes, it's just printing, e.g.,

X_new.iloc[0, 0]

gives

0    1
1    4
Name: 0, dtype: int64

and

X_new.iloc[0, 0].shape

is

(2,)

anyway, nested_univ is dodgy anyway, not just in printing, I recommend not using it.

[weenerplasticsgroup]

ok. there is no issue