Fastest way to convert 29th February into 1st March in R?

I have a vector of dates ("YYYY-MM-DD") in R, which contain some 29th February. For simplicity of my analysis, I want to convert these dates into 1st March. I am doing this using a for loop, but I wondered if there is a faster way as I will apply it to tens of thousands of dates.

Here is a reproducible example:

library(lubridate)
dates0 <-  as.POSIXlt(as.Date(c("1945-04-09 UTC", "1957-06-10 UTC", "1924-02-28 UTC",
             "1921-06-22 UTC","1926-05-16 UTC", "1920-02-29 UTC")))

dates0[6]

for(i in 1:6){
  if(day(dates0[i])==29 & month(dates0[i])==2) day(dates0[i]) <- 1; month(dates0[i]) <- 3
}

dates0[6]

Try

tmp=month(dates0)==2 & day(dates0)==29
dates0[tmp]=dates0[tmp]+days(1)

A more onerous method than user2974951 offered, but more easily generalised if the date change is more than a single day step:

tochange <- day(dates0) == 29 & month(dates0) == 2
month(dates0[tochange]) <- 3
day(dates0[tochange]) <- 1

In terms of fastest:

dates1 <- rep(dates0, times = 1000)

OP <- \(x) {
  for(i in 1:length(x)){
    if(day(x[i])==29 & month(x[i])==2) day(x[i]) <- 1; month(x[i]) <- 3
  }
}

setBoth <- \(x) {
  tochange <- day(x) == 29 & month(x) == 2
  month(x[tochange]) <- 3
  day(x[tochange]) <- 1
}

incrementDay <- \(x) {
  tmp=month(x)==2 & day(x)==29
  x[tmp]=x[tmp]+days(1)
}

microbenchmark(
  OP           =  OP(dates1),
  setBoth      =  setBoth(dates1),
  incrementDay =  incrementDay(dates1)
)

Unit: milliseconds
         expr    min      lq     mean  median     uq      max neval cld
           OP 1.1611 1.38955 1.839490 1.67015 1.8911   8.3134   100   a
      setBoth 1.2863 1.44650 3.456468 1.63335 1.7768 158.1885   100   a
 incrementDay 1.1924 1.41305 1.765843 1.53635 1.7587   7.8440   100   a

There doesn't seem to be a lot in it, my way likely slowest.

In case there is no need to use POSIXlt it would gain much speed when using Date, where, when found the 29 Feb., simply 1 needs to be added and this works in base.

i <- which(format(dates1, "%m%d") == "0229")
`[<-`(dates1, i, value = dates1[i] + 1) }

Data

dates1 <- as.Date(c("1945-04-09 UTC", "1957-06-10 UTC", "1924-02-28 UTC",
                    "1921-06-22 UTC","1926-05-16 UTC", "1920-02-29 UTC"))
dates0 <- as.POSIXlt(dates1)

Benchmark

incrementDay <- \(x) {   #@user2974951
  tmp <- month(x)==2 & day(x)==29
  x[tmp] <- x[tmp]+days(1)
  x }

idLub <- \(dates1) {
    i <- which(month(dates1)==2 & day(dates1)==29)
    `[<-`(dates1, i, value = dates1[i] + 1) }

idBase <- \(dates1) {
    i <- which(format(dates1, "%m%d") == "0229")
    `[<-`(dates1, i, value = dates1[i] + 1) }

bench::mark(incrementDay =  incrementDay(dates1),
            idLub =  idLub(dates1),
            idBase =  idBase(dates1) )

Result

 expression     min  median `itr/sec` mem_alloc `gc/sec` n_itr  n_gc total_time
  <bch:expr> <bch:t> <bch:t>     <dbl> <bch:byt>    <dbl> <int> <dbl>   <bch:tm>
1 increment… 441.3µs 454.7µs     2161.      328B     14.4  1048     7      485ms
2 idLub       46.2µs  48.8µs    20060.    21.8KB     21.0  9531    10      475ms
3 idBase      28.8µs  30.9µs    31793.        0B     22.3  9993     7      314ms

Using Date instead of POSIXlt decreases the time in this case by a factor of 10.