Duplicate items on Panel supported?

[dalejung]

import pandas as pd
import numpy as np

data = np.random.randn(5, 100, 5)
panel = pd.Panel(data, items=list("AACDE"))

panel.iloc[0] # ValueError: Wrong number of dimensions

fr = panel.loc["E"]
fr.values.shape # (5, 100, 5) wrong

So this is another contrived example that I got from just running a battery of tests on random objects. Basically having a duplicate item key will bork the Panel.

I'm actually not sure whether Panel is meant to be a collection of uniquely identified DataFrames, or more like a generic 3d container that has convenient semantics for its axes. I think at one point it was the former, but as more things pull from generic it's becoming the later?

[jreback]

should work...

[dalejung]

@jreback Also should Panel support non-scalar arith methods like: panel * panel? I hear mention of an arith refactor and wasn't sure if that was included.

[jreback]

that's a bit more complicated as it involves panel alignment (which is only partially done)

[jreback]

@dalejung I take it back, I think you can just remove that test and it should work (but needs some tests)....want to do a PR?

def _arith_method(func, name):
    # work only for scalars

    def f(self, other):
        #if not np.isscalar(other):
        #    raise ValueError('Simple arithmetic with %s can only be '
        #                     'done with scalar values' % self._constructor.__name__)

        return self._combine(other, func)
    f.__name__ = name
    return f

then this works

In [20]: p = tm.makePanel()

In [21]: tm.assert_frame_equal(2*p['ItemA'],(p + p.loc[['ItemA']])['ItemA'])

[jreback]

let me do a quick PR to fix something related......

[jreback]

jreback@adf9d48

if you would like to add some tests and do a PR....I think it will pretty much work (forgot had written the panel combiner)