Feature Request: Support indexing into python objects

[schlichtanders]

I am using pandas with PyCall.jl and would like to write something like

mysubset = df[df["colname"] == "value"]

looking into the code
https://github.com/JuliaPy/PyCall.jl/blob/master/src/PyCall.jl#L349-L361
it seems that this is just not supported currently (because of some legacy where indexing was used instead of dot access)

[schlichtanders]

I can imagine something like

Base.getindex(o::PyObject, s) = py"$o[$s]"
Base.setindex!(o::PyObject, v, s) = py"$o[$s] = $v"

[stevengj]

In the meantime, you can use PythonCall.jl, which supports this.

[schlichtanders]

This is not completely true. PythonCall.jl has indeed difficulties with the mentioned example because of comparisons between python object (series) and julia object ("value"). you need to use the @py macro if you want this to work with PythonCall.jl to the best of my knowledge.

I rather prefer to use native julia syntax, which is indeed the main reason I am going for PyCall.jl when calling Python from Julia.

[fabiensatalia]

Came here to request the same, in fact I would be happy to "just" be able to select rows from a pandas data frame, e.g. df[[1,2]] to select the first two rows