BUG: Index removes name when sliced

[cpcloud]

closes #4226

[jreback]

can u make sure no perf hit with a test_perf ?

[cpcloud]

running now

[jreback]

whenever I touch indexing I make sure to do that
seemingly innocuous changes sometimes have impact ( that said there IS some variation in perf stats so even a 15% reduction may not mean much and could change in retest)

[cpcloud]

good to know. seems like only a small constant will be added here since getslice just passes to getitem, but i'll be sure to do that whenever i touch indexing

[cpcloud]

i haven't really had time to play with test_perf.sh but is there a way to look at all test_perf.sh runs one has done? e.g., i could run overnight 1000 times and look at variability to get a reasonable estimate

[cpcloud]

well this is annoying

index_datetime_union                         |  51.5563 |  10.3513 |   4.9806 |
index_datetime_intersection                  |  51.4587 |  10.3063 |   4.9929 |
-------------------------------------------------------------------------------
Test name                                    | head[ms] | base[ms] |  ratio   |
-------------------------------------------------------------------------------

[cpcloud]

twice in a row it's been around ~5

[cpcloud]

@jreback what to do here...__getslice__ is defined by ndarray in C code so...gonna be a little hard to get around this.

[jreback]

array_finalize should set name (I thought it did)?
this is called on array creation after a numpy func

[cpcloud]

hm ok my fix looks unecessary, there's something weird in DatetimeIndex since __array_finalize__ is not being called

[cpcloud]

oh nvm i c. no name attr there nice, easy fix

[cpcloud]

perf back to fluctuating around 1.0! yay!

[cpcloud]

@jreback any more thoughts here?

[jreback]

so this actuallly ends up being a trivial change....ok...merge away