YearWeek and scale_x_yearweek with 1 year gaps lead to wrong x axis.

[USMortality]

YearWeek and scale_x_yearweek with 1 year gaps, lead to wrong x axis. See here in this example, 2020 is duplicated twice and every year, thereafter is offset by 1...

I think this has todo with the extra week 53 of 2020...
What can I do to fix this?

df <- as_tibble(data.frame(
  x = seq(from = mdy("1/1/2019"), to = mdy("1/1/2022"), by = "week"),
  y = runif(157, min = 0, max = 25)
)) %>% mutate(x = yearweek(x))

ggplot(df, aes(x = x, y = y)) +
  geom_line(color = "#5383EC", linewidth = 1) +
  scale_x_yearweek(date_breaks = "1 year", date_labels = "%Y")

[USMortality]

Workaround, use scale_x_date(date_labels = "%Y", breaks = "1 year") works fine.

df <- as_tibble(data.frame(
  x = seq(from = mdy("1/1/2019"), to = mdy("1/1/2022"), by = "week"),
  y = runif(157, min = 0, max = 25)
)) %>%
  mutate(x = yearweek(x)) %>%
  mutate(date = date(x))

ggplot(df, aes(x = date, y = y)) +
  geom_line(color = "#5383EC", linewidth = 1) +
  scale_x_date(date_labels = "%Y", breaks = "1 year")

So IMO this shows, that there's clearly a bug in the scale_x_yearweek function. Using '52 weeks' as a workaround, appears to make it work, but obviously results in wrong ticks, as there are 53 weeks sometimes...

[mitchelloharawild]

What's happening here is that an ISO week-numbering year is slightly offset from the usual Gregorian year. When you use breaks = "1 year" you are using a time duration of 1 Gregorian year between breaks, not an ISO week-numbering year.

If you use date_labels = "W%V %G", you will see the appropriate ISO week date labels.
%V is the week of the year, and %G is the ISO week-numbering year.

The second 2020 (2021-01-01 in the Gregorian calendar) is actually the 53rd week of 2020 in the ISO week calendar.

library(tsibble)
library(ggplot2)
library(dplyr)
library(lubridate)
df <- as_tibble(data.frame(
  x = seq(from = mdy("1/1/2019"), to = mdy("1/1/2022"), by = "week"),
  y = runif(157, min = 0, max = 25)
)) %>% mutate(x = yearweek(x))

ggplot(df, aes(x = x, y = y)) +
  geom_line(color = "#5383EC", linewidth = 1) +
  scale_x_yearweek(date_breaks = "1 year", date_labels = "W%V %G")

^{Created on 2026-02-16 with reprex v2.1.1}

Unfortunately date break sequencing by ISO week-numbering years is not implemented in ggplot2 (or actually scales -> lubridate). I'm currently working on fixing this in ggtime and mixtime.

As you suggest, converting yearweek into Date is a simple fix that avoids the intricacies of the ISO week calendar. Although the plot breaks will be slightly off from the data, since the 2021 break-point will be at 2021-01-01 rather than the first week of the 2021 (which is 2021-01-04 for weeks starting Monday).