Count with conditions in R dataframe - r

Question

I have the following DF:
Week SKU Discount(%)
1 111 5
2 111 5
3 111 0
4 111 10
1 222 0
2 222 10
3 222 15
4 222 20
1 333 5
2 333 0
3 333 0
I would like to have this outcome:
Week SKU Discount(%) Duration LastDiscount
1 111 5 2 0
2 111 5 2 0
3 111 0 0 0
4 111 10 1 2
1 222 0 0 0
2 222 10 3 0
3 222 15 3 0
4 222 20 3 0
1 333 5 1 0
2 333 0 0 0
3 333 0 0 0
Duration is the number of weeks that 1 SKU had discounts continuously.
LastDiscount counts the number of weeks from the last time the SKU was on a continuous discount, only if there are weeks with 0 in between discounts.

Answer 1

One option to check the "Duration' is after grouping by 'SKU', use rle (run-length-encoding) on a logical vector, gets the lengths and 'values' and replicate those duration. Similarly, the "LastDiscount" can be obtained by getting the sum of logical values
library(dplyr)
df1 %>%
group_by(SKU) %>%
mutate(Duration = with(rle(Discount > 0), rep(lengths*values,
lengths)),
temp = with(rle(Discount > 0), sum(values != 0)),
LastDiscount = if(temp[1] > 1) c(rep(0, n()-1), temp[1]) else 0) %>%
select(-temp)
# A tibble: 11 x 5
# Groups: SKU [3]
# Week SKU Discount Duration LastDiscount
# <int> <int> <int> <int> <dbl>
# 1 1 111 5 2 0
# 2 2 111 5 2 0
# 3 3 111 0 0 0
# 4 4 111 10 1 2
# 5 1 222 0 0 0
# 6 2 222 10 3 0
# 7 3 222 15 3 0
# 8 4 222 20 3 0
# 9 1 333 5 1 0
#10 2 333 0 0 0
#11 3 333 0 0 0
Or using data.table
library(data.table)
i1 <- setDT(df1)[, grp := rleid(Discount > 0), SKU][Discount > 0,
Duration := .N, .(grp, SKU)][,
LastDiscount := uniqueN(grp[Discount > 0]), .(SKU)][,
tail(.I[Discount > 0 & LastDiscount > 1], 1), SKU]$V1
df1[-i1, LastDiscount := 0][]
data
df1 <- structure(list(Week = c(1L, 2L, 3L, 4L, 1L, 2L, 3L, 4L, 1L, 2L,
3L), SKU = c(111L, 111L, 111L, 111L, 222L, 222L, 222L, 222L,
333L, 333L, 333L), Discount = c(5L, 5L, 0L, 10L, 0L, 10L, 15L,
20L, 5L, 0L, 0L)), class = "data.frame", row.names = c(NA, -11L
))