Overload df.getitem with idx of tuple

[densmirn]

No description provided.

[kozlov-alexey]

I think we should state this as a limitation in our documentation too.
Also it might look better if you write it this way:

        test_impl = lambda df: df[['A', 'C']]
        sdc_func = self.jit(lambda df: df[('A', 'C')])

or use do_jit parameter, e.g.

      # Pandas implementation does not support idx as a tuple
       # but SDC implementation only supports tuple, so adjust tested code
       def test_impl(df, do_jit):
           if do_jit == True:  # noqa
               return df[('A', 'C')]
           else:
               return df[['A', 'C']]
       sdc_func = self.jit(test_impl)

       pd.testing.assert_frame_equal(sdc_func(df, True), test_impl(df, False))

[kozlov-alexey]

The exception should really have a message, maybe something like "Column is not in the DataFrame" would fit?

[kozlov-alexey]

This looks as a very generic function, that we could use everywhere else, so it might be useful to move it to sdc_typing_utils. So it's responsibility would be to call exec on provided func_code and return back created func object.

[kozlov-alexey]

This should probably be a more strict check, e.g.

Suggested change

	if isinstance(idx, types.Tuple):
	if (isinstance(idx, types.Tuple)
	and all([isinstance(item, types.StringLiteral) for item in idx])):

[kozlov-alexey]

Looks very good! No major comments from me, only minor ones, that might be applied later on.