reconverse / grates Goto Github PK

Add an example of using a date for the offset in as_period()

Current implementations are very minimal and leave a lot to be desired (e.g. #17). They really should be classed as "experimental" until developed more.

add wrappers around as_yearweek for epiweeks and isoweeks

I.e. the following should not error

library(grates)
yearweek(2002L,1:2)
#> Error in yearweek(2002L, 1:2): `year` and `week` must be the same length.

^{Created on 2023-03-02 with reprex v2.0.2}

Should just be copy of the x scales. Related to #20 .

• Perhaps [start,finish] instead of start to finish
• Label is not used in period scale so this would still look ok

This would be useful for downstream {incidence2} package but it may be better to just use IDates from {data.table}.

The current constructors are a little weird in that they want stored as counts from the unix epoch which is the internal representation of grates object. Perhaps better constructors would allow more intuitive specification:

instead of

isoweek(0L)

it would be nice to write

isoweek(year = 1970, week = 1)

perhaps the current constructors would be provided as minimal low level ones and renamed with a new_ prefix. This way they are still provided in case useful to other developers. E.g.

new_isoweek(0L)

Would be great if we could specify the position of breaks in grates::scale_x_grates_isoweek(breaks = ...)

I've just noticed these functions are missing some arguments. These calls probably want replacing with bespoke abort() but then need to ensure these error as desired.

Tests are currently failing on GitHub actions for the the old release version - R 4.1.3 (2022-03-10) on Ubuntu.

Am I using any new functionality somewhere in my tests or is something else causing this???