[jira] [Commented] (ARROW-12637) Range type (anchored variable-length interval)

["Joris Van den Bossche (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ARROW-12637?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17339231#comment-17339231 ] 

Joris Van den Bossche commented on ARROW-12637:
-----------------------------------------------

What do you mean with fixed length? Each interval/range having the same length? That's not a restriction in pandas, my simple example just had regular ranges (pandas stores left/right in separate arrays, very similar as a struct array).

> Range type (anchored variable-length interval)
> ----------------------------------------------
>
>                 Key: ARROW-12637
>                 URL: https://issues.apache.org/jira/browse/ARROW-12637
>             Project: Apache Arrow
>          Issue Type: Wish
>          Components: C, C++
>            Reporter: Jacek Pliszka
>            Priority: Major
>
> This is wish for range(interval) extension type.
>  
> Type would describe range/interval with defined start and end and would have metadata
> whether range is open/close/left-open/right-open.
>  
> Considerations:
>  * should it focus on integers and like or on floats as well?
>  * are open/semi-open cases worth the effort? Maybe not.
>  
> The most useful case (big low hanging fruit) might be closed range of 32bit/64bit and maybe 8bit/16bit unsigned integers packed in twice larger integer allowing usage with date/time via ordinals.
>  
> This would allow computations like:
>  
>  * overlap joins (see R foverlaps)
>  * range algebra: unions, intersections, differences
>  * classification in Allen's interval algebra
>  * conversion to/from 2 columns/rows
>  * exploding (see Pandas explode)
>  
>  
>  



--
This message was sent by Atlassian Jira
(v8.3.4#803005)