Turn a loop based code into a vectorised one in R? - r

Question

I´ve got this dataset and want to perform some calculations based on certain conditions:
library(tidyverse)
library(lubridate)
filas <- structure(list(Año = c(rep(2020,4),rep(2021,4),2022),
Mes = c(2:5,3:4,9,11,1),
Id = c(rep(1,7),2,2)),
row.names = c(NA, -9L),
class = c("tbl_df", "tbl", "data.frame")) %>%
mutate(fecha = make_date(Año,Mes,1),
meses_imp = make_date(2999,1,1))
Año
Mes
Id
fecha
meses_imp
2020
2
1
2020-02-01
2999-01-01
2020
3
1
2020-03-01
2999-01-01
2020
4
1
2020-04-01
2999-01-01
2020
5
1
2020-05-01
2999-01-01
2021
3
1
2021-03-01
2999-01-01
2021
4
1
2021-04-01
2999-01-01
2021
9
1
2021-09-01
2999-01-01
2021
11
2
2021-11-01
2999-01-01
2022
1
2
2022-01-01
2999-01-01
I need to add rows for each "Id" when there are "holes" between two consecutive ones, and count those added rows later. I´ve achieved this using a "while" loop:
i <- 2
while(!is.na(filas[i,]$Id)) {
if (as.double(difftime(filas[i,]$fecha,filas[i-1,]$fecha)) > 31 &
filas[i,]$Id == filas[i-1,]$Id) {
filas <- add_row(filas,
Id = filas[i,]$Id,
fecha = filas[i-1,]$fecha + months(1),
meses_imp = pmin(filas[i-1,]$fecha,
filas[i-1,]$meses_imp),
.after = i-1)}
i=i+1}
filas2 <- filas %>%
group_by(Id,meses_imp) %>%
summarise(cant_meses_imp = n()) %>%
ungroup() %>%
filter(meses_imp != "2999-01-01")
filas <- left_join(filas,
filas2,
by=c("Id","meses_imp"))
Año
Mes
Id
fecha
meses_imp
cant_meses_imp
2020
2
1
2020-02-01
2999-01-01
NA
2020
3
1
2020-03-01
2999-01-01
NA
2020
4
1
2020-04-01
2999-01-01
NA
2020
5
1
2020-05-01
2999-01-01
NA
NA
NA
1
2020-06-01
2020-05-01
9
NA
NA
1
2020-07-01
2020-05-01
9
NA
NA
1
2020-08-01
2020-05-01
9
NA
NA
1
2020-09-01
2020-05-01
9
NA
NA
1
2020-10-01
2020-05-01
9
NA
NA
1
2020-11-01
2020-05-01
9
NA
NA
1
2020-12-01
2020-05-01
9
NA
NA
1
2021-01-01
2020-05-01
9
NA
NA
1
2021-02-01
2020-05-01
9
2021
3
1
2021-03-01
2999-01-01
NA
2021
4
1
2021-04-01
2999-01-01
NA
NA
NA
1
2021-05-01
2021-04-01
4
NA
NA
1
2021-06-01
2021-04-01
4
NA
NA
1
2021-07-01
2021-04-01
4
NA
NA
1
2021-08-01
2021-04-01
4
2021
9
1
2021-09-01
2999-01-01
NA
2021
11
2
2021-11-01
2999-01-01
NA
NA
NA
2
2021-12-01
2021-11-01
1
2022
1
2
2022-01-01
2999-01-01
NA
Since I`d like to apply this to a much larger dataset (~ 300k rows), how could I rewrite it in a vectorised way so it´s more efficient (and elegant maybe)?
Thanks!

Answer 1

You can apply the following code using padr and zoo packages.
This idea is to:
Add missing dates with the padr::pad() function.
Remove unwanted lines (non-integer Id values)
Create na and grp columns to identify rows added in 1.
Group by grp and create a column cant_meses_imp to count the number of consecutive na in each group
Select only desired columns
library(dplyr)
library(padr)
library(zoo)
filas %>%
pad(by = "fecha") %>% # add missing dates
mutate(Id = na.approx(Id)) %>% # interpolate NA values in Id column
subset(Id%%1 == 0) %>% # Keep only Id interger
# This part is for generating the cant_meses_imp column
mutate(na = ifelse(is.na(Mes), 1, 0),
grp = rle(na)$lengths %>% {rep(seq(length(.)), .)}) %>%
group_by(grp) %>%
mutate(cant_meses_imp = ifelse(na == 0, NA, n())) %>%
ungroup() %>%
select(-c(na, grp))
The code does not reproduce exactly the fecha column as there is no guidelines for its values.