ENH: make Series.ptp ignore NaN values (GH11163)

[ajcr]

Closes GH11163.

Series.ptp will now ignore missing values when calculating the peak to peak difference.

[jreback]

actually what I would is define this using generic.py/_make_stat_function (e.g. how .min and such are defined). though this would only apply to Series (so might need to move a couple of things around to deal with this).

[ajcr]

Ah OK - no problem, I should be able to take a look at this and update in the next week or so.

[jreback]

move to 0.17.1

[jreback]

put in the API section as we are going to change the signature a bit

[jreback]

can you update

[ajcr]

Apologies - the past few weeks have been much busier than I'd anticipated. I hope to get back to update this very soon.

[jreback]

no prob.

[jreback]

@ajcr progress?

[jreback]

@ajcr pls update when you can

[ajcr]

@jreback: my apologies for the delay once again, I've just got back and was looking at this again just now.

As I understand it, I need to do the following:

• make a nanptp function in nanops.py (is implementing this as nanmax - nanmin sufficient?)
• add an entry for ptp to _make_stat_function in generic.py
• make sure that this doesn't become a method of DataFrame - i.e. retrict it to Series

The last part I'm not quite sure how to do best, because NDFrame is a parent class of both Series and DataFrame. Are there any other methods from NDFrame that are restricted to one or the other object in this way?

[jreback]

you don't really need to do step 1, simply define it in _make_stat_function to do this (as its sort of a trivial function)

if you look at the end of core/series.py there is a like Series._add_numeric_operations(), (and in DataFrame), so you just need to tag certain of these functions to be added only on certain classes.

[ajcr]

@jreback: made some changes and things are now green - let me know if you'd like me to improve anything or make any other changes.

[jreback]

thanks!

[ajcr]

No problem (sorry it took a while) - thanks for your guidance!