Make round_date() really use week_start=

[xvrdm]

After reading the documentation and many SO questions, I still couldn't really explain the difference between round_date() and ceiling/floor_date when using unit="week" and week_start=1.

I thought it might just be that round_date() ignore week_start=, a bit like floor/ceiling_date back in the #509 days.

Without week_start=, everything looks as expected.

> date <- parse_date_time("November 27 2018 23:45", orders="bdyHM")
> date
[1] "2018-11-27 23:45:00 UTC"
> lubridate::round_date(date, "week")
[1] "2018-11-25 UTC"
> lubridate::floor_date(date, "week")
[1] "2018-11-25 UTC"
> lubridate::ceiling_date(date, "week")
[1] "2018-12-02 UTC"

But if you ask for weeks starting on Mondays (or any other day). Only floor/ceiling_date seem affected:

> lubridate::round_date(date, "week", week_start = 1)
[1] "2018-11-25 UTC"
> lubridate::floor_date(date, "week", week_start = 1)
[1] "2018-11-26 UTC"
> lubridate::ceiling_date(date, "week", week_start = 1)
[1] "2018-12-03 UTC"

Apart from this tiny glitch, thanks for the awesome library: I don't want to use anything else when it comes to dates 👍 !

[xvrdm]

Closing this for now. I tried to add tests in the style of ceiling_date()/floor_date():

# in test-round.r
  expect_equal(wday(round_date(ct, "week", week_start = 1)), 2)
  expect_equal(wday(round_date(ct, "week", week_start = 2)), 3)
  expect_equal(wday(round_date(ct, "week", week_start = 5)), 6)
  expect_equal(wday(round_date(ct, "week", week_start = 7)), 1)
  expect_equal(wday(round_date(date, "week", week_start = 1)), 2)
  expect_equal(wday(round_date(date, "week", week_start = 2)), 3)
  expect_equal(wday(round_date(date, "week", week_start = 5)), 6)
  expect_equal(wday(round_date(date, "week", week_start = 7)), 1)

  expect_equal(wday(round_date(ct, "week", week_start = 1), week_start = 1), 1)
  expect_equal(wday(round_date(ct, "week", week_start = 2), week_start = 2), 1)
  expect_equal(wday(round_date(ct, "week", week_start = 5), week_start = 5), 1)
  expect_equal(wday(round_date(ct, "week", week_start = 7), week_start = 7), 1)
  expect_equal(wday(round_date(date, "week", week_start = 1), week_start = 1), 1)
  expect_equal(wday(round_date(date, "week", week_start = 2), week_start = 2), 1)
  expect_equal(wday(round_date(date, "week", week_start = 5), week_start = 5), 1)
  expect_equal(wday(round_date(date, "week", week_start = 7), week_start = 7), 1)

Two don't pass yet.

test-round.R:484: failure: round on week respects week_start
wday(round_date(ct, "week", week_start = 5)) not equal to 6.
1/1 mismatches
[1] 1 - 6 == -5

test-round.R:493: failure: round on week respects week_start
wday(round_date(ct, "week", week_start = 5), week_start = 5) not equal to 1.
1/1 mismatches
[1] 3 - 1 == 2

Will try to do more testing.

[vspinu]

Only floor/ceiling_date seem affected:

Both ceiling_date and floor_date work as expected in your example. It's really only the issue of round_date, looks like.

[xvrdm]

Thanks, that is what I meant but apparently my brain did not work correctly :)... Corrected now.

[xvrdm]

Actually now I remember. By "affected" I meant that the week_start argument has the expected effect on them, but not on round_date. Sorry, it might be poor english...

[vspinu]

Fixed in master. Your fix was almost there it's just that week_start is not the 3rd arg in ceiling_date.

[xvrdm]

Thanks! I guess that will teach me to be more careful before searching complicated explanations!!

[StatsRhian]

Thanks for raising this @xvrdm. Was the PR ever merged? lubridate::round_date(date, "week", week_start = 1) is still ignoringweek_start for me in the latest development version of the package from GitHub.

[xvrdm]

I thought it was because I made a mistake in my PR that was kindly fixed by @vspinu . But I have not tested it in a while.

[vspinu]

Yes. It's fixed, but not on CRAN yet.

[khayes888]

It seems this issue has still not been fixed on CRAN.

round_date(ymd("2019-10-22"),unit="week",week_start = 5)
[1] "2019-10-20"
round_date(ymd("2019-10-22"),unit="week",week_start = 1)
[1] "2019-10-20"
sessionInfo()
R version 3.5.1 (2018-07-02)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows >= 8 x64 (build 9200)

Matrix products: default

locale:
[1] LC_COLLATE=English_United States.1252 LC_CTYPE=English_United States.1252 LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C LC_TIME=English_United States.1252

attached base packages:
[1] stats graphics grDevices utils datasets methods base

other attached packages:
[1] lubridate_1.7.4

loaded via a namespace (and not attached):
[1] Rcpp_0.12.18 crayon_1.3.4 dplyr_0.7.6 assertthat_0.2.0 R6_2.2.2 magrittr_1.5 pillar_1.3.0 stringi_1.2.4
[9] rlang_0.2.2 rstudioapi_0.7 bindrcpp_0.2.2 tools_3.5.1 stringr_1.3.1 glue_1.3.0 purrr_0.2.5 yaml_2.2.0
[17] compiler_3.5.1 pkgconfig_2.0.2 bindr_0.1.1 knitr_1.20 tidyselect_0.2.4 tibble_1.4.2