epi_slide and epix_slide args

[dajmcdon]

I still find before/after confusing. It takes mental math and careful reading of the documentation. It's more general, but not really how our target users are likely to think.

I suggest we bring back n for the window size as well as align = c("left", "center", "right"). We can keep before/after as NULL and ensure there aren't any conflicts. Then immediately process n+align into the appropriate before+after pair.

[brookslogan]

Thanks for the ping. I think this fell down the stack due to some engineering priorities + new testers getting tripped up by other things. Pinging @dshemetov here in case he's interested in tackling; seems related to time type rework. [Otherwise I can take a look after my other currently-planned slide edits.]

@dajmcdon have you ever used / remember us using something that's not left/center/right aligned? I know we've used non-odd-n center-aligned things at least (e.g., before = 6, after = 7), but that's the closest I can think of.

Context/history for implementer:

• We didn't move to before & after for generality; epi_slide actually had an n + align + before [with the latter two mutually exclusive]; this allowed for the same level of generality as before + after. The actual reason for the change had to deal with [epix_slide stuff; described at end] + trying to mirror interfaces between epix_slide and epi_slide; the latter's no longer a key goal.
  • Plan since this summer has been to urgently change back to n + align + before.
  • Though n + before seems awkward; I like @dajmcdon's suggestion of n + align + before + after.... assuming we want to support this level of generality at all.
• We should make sure to mirror @dshemetov's time type type rework rules when implementing this change.
• (History: how did epix_slide make [us] change epi_slide? n was always misleading in epix_slide to some people (and sometimes misleading in epi_slide, though mirroring the recent time type rework should take care of some cases [without needing time_step / allowing silently-buggy/confusing alternatives]; the remaining cases are the continually-frustrating/complicated cases with [edges and gaps of the data set, and] with non-epikey-grouped epi_slides with differing epikey availability), so we changed to epix_slide(before =) and, for consistency, epi_slide(before =, after =).)

[dshemetov]

Happy to tackle this, should be pretty simple. I don't have strong opinions on the interface here, it's all equivalent to me.

[dshemetov]

See #513 for a simple prototype.