Implement "nlargest()", "nsmallest()" methods.

[garciparedes]

It would be interesting to implement this kind of methods over the DataFrame class. I think the needed memory to compute that values shouldn't be so big and could be really interesting to provide insightful indicators.

Here is the pandas counterpart documentation:

• nlargest: https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.nlargest.html
• nsmallest: https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.nsmallest.html

[maartenbreddels]

Hi Sergio,

yes, I think we can do this. It's a bit tricky with how vaex works internally, but I'll keep this in mind when doing some refactor work. Let's keep this issue open as a reminder.

cheers,

Maarten

[garciparedes]

Hi @maartenbreddels,

Thank you so much for accepting this feature request. 🙂

If you don't mind, I've a doubt about why the vaex implementation of this kind of statistics would be a bit tricky. Is the reason related with the "vectorial" nature of the output?

[maartenbreddels]

it has to do with how vaex filters. Vaex always works with the unfiltered raw data, which means it is always tricky to map between an unfiltered index (say the 6th element of the unfiltered array) and the filtered index.