Missing ellipsis cells in pivot_table.to_html when limited by max_cols and len(rows) > 1

[BertrandBordage]

import pandas
df = pandas.DataFrame([
    {'a': 'aa', 'b': 'ba', 'c': 'ca', 'd': 1},
    {'a': 'ab', 'b': 'bb', 'c': 'cc', 'd': 2},
    {'a': 'ac', 'b': 'bc', 'c': 'cc', 'd': 3}
])

df.pivot_table(cols=['a'], rows=['b', 'c']).to_html(max_cols=2)

This renders like this:

while this:

df.pivot_table(cols=['a'], rows=['b']).to_html(max_cols=2)

renders like this:

As you can see in the first table, there are no ellipsis cells like in the second one.
I inspected the code to find the reason for this bug but couldn't find…

[mroeschke]

This looks fixed; output formatting for dataframes has changed since this version though. A test would be good to confirm

[simonjayhawkins]

using pandas 0.23.0 i appear to get the desired behaviour

however, i came to the issue from #15019, so decided to add index=False to the output

[mroeschke]

If you're up for contributing a test @simonjayhawkins to confirm this behavior, it would be much appreciated!

[simonjayhawkins]

the column misalignment in Out[4] and Out[5] above is a separate issue. This was fixed for a column MultiIndex in #8452. adding an index MultiIndex to the same example appears to still gives the column offset bug.

so this should probably be raised as a new issue?

[mroeschke]

Sure thing; you can open a new issue. Please use code snippets and not pictures of code as it's harder to reproduce.

[simonjayhawkins]

Opened #22579 for the column misalignment.

[simonjayhawkins]

@mroeschke currently in pandas\tests\io\formats\test_to_html.py there is a test:

    def test_to_html_truncate_multi_index(self):
        pytest.skip("unreliable on travis")
        arrays = [['bar', 'bar', 'baz', 'baz', 'foo', 'foo', 'qux', 'qux'],
                  ['one', 'two', 'one', 'two', 'one', 'two', 'one', 'two']]
        df = DataFrame(index=arrays, columns=arrays)
        fmt.set_option('display.max_rows', 7)
        fmt.set_option('display.max_columns', 7)
        result = df._repr_html_()
        expected = '''\
...

which tests that:

produces:

which I think covers this issue. so i'm thinking that adding another test may not be necessary?

[mroeschke]

Although these tests look similar, it would still be good to confirm that the to_html pathway is tested. The test uses a private method _repr_html_ which users should not normally be accessing.