Pandas .loc() function chokes while using PyCall

[milanhej]

PyCall has proven invaluable to me -- thanks! -- but I have been unable to use the important .loc() function (handy in locating specific rows and columns of a dataframe) under Pandas.

The following python works to locate two rows and a column:

import pandas as pd
data = {'Name':['Bob', 'Ruth', 'Tina', 'Jack'], 'Age':[30, 25, 18, 31]} 
df = pd.DataFrame(data)
df.loc[0:1,"Age"]

But under Julia while using PyCall, the equivalent code fails:

using PyCall
pd = pyimport("pandas")
data = Dict("Name"=> ["Bob", "Ruth", "Tina", "Jack"], "Age" => [30, 25, 18, 31] )
df = pd.DataFrame(data)
df.loc(1:2,"Age")

Thanks for any insight how to work around this issue.

[tkf]

I think we need to define getindex(::PyObject, ...) so that it just works. Until then, you need to use https://github.com/JuliaPy/Pandas.jl or just do py""" .... """.

[stevengj]

The equivalent code right now is:

get(df.loc, (0:1,"Age"))

(The get(o, key) function is the documented translation of Python's o[key] at the moment: https://github.com/JuliaPy/PyCall.jl#pyobject).

We will implement getindex eventually, but have been holding off because getindex used to be used for getting attributes in PyCall, before dot overloading. That's why it's currently deprecated. Once enough PyCall-using packages have upgraded, we can switch to making getindex in Julia be equivalent to the Python o[key].