ERR: better error message for trying to concat non-pandas objects

[jreback]

In [3]: pd.concat([Series(), 'foo'])
TypeError: cannot concatenate a non-NDFrame object

trivially non-NDFrame -> non-pandas. but I think even better would be something like

In [4]: objs = [Series(), 'foo']

In [7]: "cannot concatenate [{pandas}] with non-pandas objects of type [{non_pandas}]".format(
   ...: pandas=','.join(set([str(o.__class__.__name__) for o in objs if isinstance(o, pandas.core.generic.NDFrame)])),
   ...: non_pandas=','.join(set([str(type(o)) for o in objs if not isinstance(o, pandas.core.generic.NDFrame)])))
Out[7]: "cannot concatenate [Series] with non-pandas objects of type [<class 'str'>]"

[petioptrv]

I would like to take care of this one. There could be more than two objects in the concatenation, so how about "cannot concatenate objects of type [<class 'str'>], all objects must be pandas objects"?

[jreback]

@petioptrv yes that's why I put the list comprehension and set things. Should work with many objects.

Its just meant to be more instructive to the user.

[jorisvandenbossche]

I think "pandas objects" is still a bit vague. Eg an Index, or a Timestamp are also pandas objects, but cannot be concatenated.
Are there other things than Series, DataFrame and Panel that can be concatenated? If not, let's just list those in the error message?

eg

TypeError: cannot only concatenate with Series, DataFrame or Panel objects, an object of type [<class 'str'>] was passed.

[higs4281]

Looking at this.