BUG: bug when trying to display an embedded PandasObject (GH5324)

[jreback]

closes #5324

There was a bug where this wasn't display at all, so this PR fixes

Do we care about formatting embedded pandas objects? (as we don't recommend this anyhow)

In [15]: df1 = DataFrame({'a':[1, 2, 3], 'b':[3, 4, 5]})

In [19]: df2.dtypes
Out[19]: 
0    object
dtype: object

In [16]: df1
Out[16]: 
   a  b
0  1  3
1  2  4
2  3  5

In [17]: df2 = DataFrame([df1,df1+10])

In [18]: df2
Out[18]: 
                                         0
0             a  b
0  1  3
1  2  4
2  3  5
1      a   b
0  11  13
1  12  14
2  13  15

[jreback]

@y-p @jtratner ?

[jtratner]

👍

[jreback]

This apparently is what 0.10.0 displayed, which is wrong (because it appears they are lists), but then again
does show a reasonable display.....

        0
0  [a, b]
1  [a, b]
>>> 

[jreback]

to contrived?

In [4]: DataFrame([ DataFrame({'A' : [1,2], 'B' : 0}), Series([1,2,3]) ])
Out[4]: 
           0
0  DataFrame
1     Series

[jtratner]

you should just use numpy's array_repr for Series

[ghost]

I'd say this falls under "undefined behaviour". No objections to making this change though.