week() does not return actual calendar week

[STATWORX]

Hi,

the week() function does not return the actual calendar week (what I think most users expect).
Please fix or rename function since a lot of users are expecting a different result.

Thank you!

Example:

### Packages
library(data.table)
library(lubridate)

### Thursday, 7th january 2016, calendar week 1
my_date <- as.Date("2016-01-07", format = "%Y-%m-%d")

### Week from data.table
data.table::week(my_date)
> 2 # not what most people would expect

### Week from lubridate
lubridate::isoweek(my_date)
> 1 # as expected

[MichaelChirico]

please include the output -- I'm not at a machine so I can't tell what you see/expect.

For now, please see the relevant code and I particular the preceding comment section:

https://github.com/Rdatatable/data.table/blob/master/R/IDateTime.R

[MichaelChirico]

see also #1658 which i imagine is related

[STATWORX]

Updated the example. Actually, the week() produces this: "Weeks is the number of complete seven day periods that have occured between the date and January 1st, plus one..."(see http://stackoverflow.com/questions/27116645/r-week-function-returns-unexpected-values). I think that this is an unexpected return that a lot of users are not aware of.

[DavidArenburg]

According to you, what will most users expect for "2016-01-08"? Both lubridate and data.table will return 2, while according to this for instance, first week is until 2016-01-10. Did you actually look at the source code of data.table::week and lubridate::week? They are almost identical with a small change (which was implied by @MichaelChirico already) - both written in a very simple base R code which you could easily understand exactly what it's doing.

[STATWORX]

OK, you're right. lubridate:week() and data.table::week() are almost identical. Just checked the source code. Issue updated. We're using the lubridate:isoweek() function now for the actual calendar week.

[arunsrinivasan]

Thanks @STATWORX for filing the issue (by following up from twitter). Perhaps we'll need an isoweek as well. Good to know.

[arunsrinivasan]

@MichaelChirico do you think you've some time to look up into isoweek?

[MichaelChirico]

sure, why not =D