WISH: extract/width

[hiiamboris]

Or whatever name is better. See #94 for background, but this will be useful in any case.

Currently extract is limited to extraction of 1 item over period P, whereas it would be more powerful if we could extract N items over period P:

>> extract/width/skip [1 2 3 4 5 6 7 8] 2 4
== [1 2 5 6]

Currently supported /index refinement OTOH is not really needed as it can be done with skip (but perhaps a nice convenience).

[greggirwin]

For a name /part makes sense. No /width as that's a required parameter already. Would then also need to decide if /part and /index work together. Since extract is already a mezz, mocking this up should be easy.

[hiiamboris]

USAGE:
     EXTRACT series width

DESCRIPTION: 
     Extracts a value from a series at regular intervals. 
     EXTRACT is a function! value.

ARGUMENTS:
     series       [series!] 
     width        [integer!] "Size of each entry (the skip)."

REFINEMENTS:
     /index       => Extract from an offset position.
        pos          [integer!] "The position."
     /into        => Provide an output series instead of creating a new one.
        output       [series!] "Output series."

width here is internal name and can be changed at will. While /part has a totally different meaning: to limit the extraction region (for consistency with other series functions). Or /size if you don't like /width ;)

Ideally:

>> extract/width/part [1 2 3 4 5 6 7 8 9 10] 4 2 5
== [1 2 5]

For a mockup see #94

Yes, /width begins from /index.

[GalenIvanov]

I had a similar wish

The motivation is to have an easy way of splitting series into regular chunks (overlapping if needed).

[hiiamboris]

@galenivanov at a glance, extraction of overlapping chunks seems to be a rare use case. What do you think?

[GalenIvanov]

@hiiamboris I agree that overlaping chunks are not used very often. A possible case is a filter-like code, that needs to know the neighbours of a value (usually done for large sets of data - needs a low-level implementation). My idea is that overlapping comes at no cost, when the part argument is bigger than the skip argument.

[hiiamboris]

https://github.com/red/red/discussions/5340