BUG/TST: fix several issues with slice bound checking code

[immerrr]

This patch removes more of what's left after removing pandas-specific customizations to slice indexing.

BUG/TST: obj.iloc[:,-4:-10] should return empty slice (because slice.step > 0 and slice.start > slice.stop)
There's no way to describe this by setting stop to a value in [0; len] interval, it should remain negative (see this ticket in python bugtracker]. Proposed fix to this issue is to remove the custom logic altogether. A side-effect of that is correct handling of negative steps in presence of out-of-bounds slices, e.g. obj.iloc[:,10:4:-1] would've incorrectly returned empty slice.

BUG: fix exception raised by Series([1,2,3]).iloc[10:].
After I've cleaned up suspicious code that handled empty-sequence keys in #6440, this piece of code started causing trouble:

if start >= l:
    return self._getitem_axis(tuple(),axis=axis)

CLN: #6299 & #6443 rendered core.indexing._check_slice_bounds function useless and it's now gone. Next step would be to drop unused raise_on_error parameter in slice-related functions.

Not sure, how do I put down all these relations into changelog, advice on that is welcome.

[jreback]

you can drop the raise_on_error parameter as it was just something to 'fix' the slicing issues (from somewhere in 0.12-0.13 transition).

Just put this all in a single line in API changes (its not really an API change, but more prominent), list all of the related issues.

[immerrr]

Just put this all in a single line in API changes (its not really an API change, but more prominent), list all of the related issues.

Did I get you correctly: immerrr@c7e6345#diff-5a809a5e093be4fe1396cbadd5bbeccaR110 ?

[jreback]

tep

[jreback]

yep

[immerrr]

Ok, folding/rebasing it then.

[jreback]

minor change

I think you can move _slice from frame/panel to generic (but series still overrides)

[immerrr]

Whoops, apparently I've accidentally deleted wrong gh branch when cleaning up local ones, sorry about that.

[immerrr]

Also, _slice is now in ndarray as requested.

[jreback]

can you rebase on currentmaster and push again.thxs

[immerrr]

Ok, done.

[jreback]

thanks as always!

if you are interestdd #6536

[immerrr]

Sure, that was a low hanging fruit :)