R: creating continuous timelines - r

Question

I have some data showing the duration individuals spent in states a, b, and c. The durations should not overlap. In my data, c sometimes overlaps with a and b. I am trying to correct the data so there are no overlaps. a and b take priority, so I am only changing the start or end dates of c.
For example, my data looks like this (for two individuals):
structure(list(id = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L,
2L, 2L), period = c("a", "b", "c", "a", "b", "c", "a", "a", "b",
"c", "a", "b"), start = c(3L, 11L, 8L, 18L, 20L, 26L, 31L, 1L,
11L, 13L, 15L, 16L), end = c(10L, 12L, 20L, 19L, 25L, 32L, 35L,
10L, 14L, 17L, 15L, 20L)), class = "data.frame", row.names = c(NA,
-12L))
And the desired output would be:
In this example, the individual with id == 2 has no time in period c, because periods a and b completely overlap.
How would you do this?

Answer 1

My method is a bit crude, but it'll work
library(tidyverse)
df %>% mutate(rowid = row_number()) %>%
pivot_longer(cols = c(start, end), names_to = "event", values_to = "val") %>%
mutate(event = factor(event, levels = c("start", "end"), ordered = T)) %>%
arrange(rowid, id, period, event) %>%
mutate(val = ifelse(period == "c" & event == "start" & val < lag(val), lag(val)+1, val),
val = ifelse(period == "c" & event == "end" & val > lead(val), lead(val)-1, val)) %>%
pivot_wider(id_cols = c(rowid, id, period), names_from = event, values_from = val) %>%
arrange(rowid) %>% mutate(start = ifelse(start > end, NA, start),
end = ifelse(end < start, NA, end)) %>%
select(-rowid)
# A tibble: 12 x 4
id period start end
<int> <chr> <dbl> <dbl>
1 1 a 3 10
2 1 b 11 12
3 1 c 13 17
4 1 a 18 19
5 1 b 20 25
6 1 c 26 30
7 1 a 31 35
8 2 a 1 10
9 2 b 11 14
10 2 c NA NA
11 2 a 15 15
12 2 b 16 20
Adding - instead of NA will turn the column type from dbl to chr.
EDIT on second thoughts I propose smaller code
df %>% group_by(id) %>%
mutate(start = ifelse(period == "c" & start < lag(end), lag(end)+1, start),
end = ifelse(period == "c" & end > lead(start), lead(start)-1, end),
start = ifelse(start > end, NA, start),
end = ifelse(start > end, NA, end))
# A tibble: 12 x 4
# Groups: id [2]
id period start end
<int> <chr> <dbl> <dbl>
1 1 a 3 10
2 1 b 11 12
3 1 c 13 17
4 1 a 18 19
5 1 b 20 25
6 1 c 26 30
7 1 a 31 35
8 2 a 1 10
9 2 b 11 14
10 2 c NA NA
11 2 a 15 15
12 2 b 16 20