iloc errors when end is greater than length

[jseabold]

IMO this should work

x = pd.DataFrame(np.random.rand(5,3))
x.iloc[:6]

Consider

np.arange(5)[:7]

And

range(5)[:7]

I'm on

[~/]
[14]: pd.version.version
[14]: '0.13.1-63-gcc6ee40'

[jreback]

see very end of the section: http://pandas-docs.github.io/pandas-docs-travis/indexing.html#selection-by-position

I believe this originally (with iget/icol) was done on purpose. e.g. let the user know.

you are advocating to allow it?

FYI, you can use head/tail to avoid this problem as they won't run off the end

[jseabold]

Yeah I'm trimming based on something else that's not a pandas object (because rolling_apply, etc. work with arrays only for now as remarked elsewhere), so I have something like

x.iloc[:len(y)]

I worry about alignment later. I just expected it to work since it's common that it works elsewhere in numpython. E.g.,

range(5)[7:9]

User's problem to check for emptiness.

[jreback]

ok. easy enough to do....I think that's fine

[jreback]

@jseabold this actually helped with another issue on multi-indexing...(with unbalanced levels)....

anyhow...take a look at #6299....see if that looks reasonable