How to improve a for loop with selection

[AndrePaind]

Good afternoon guys,
I'm trying to improve the "for" loop I've just written.
Checking online seems I have to use "sapply" instead of the "for" loop, but i don't know how to write it.

How can I improve this loop? considering length(A) is more than 800.000 elements

for(i in seq_along(a)) {
  ten_ <- AD$rain[i:(i+1)]
  tene <- sum(ten_)
  tenemean <- tene / 2
  ten[i] <- tenemean
  ten[is.na(ten)]<-0
}

thanks a lot guys, i'm still learning

[nirgrahamuk]

AD <- list()
AD$rain <-1:10
a <- 1:9
ten <- rep(NA,9)

for(i in seq_along(a)) {
  tene <- sum(AD$rain[i:(i+1)])
  tenemean <- tene / 2
  ten[i] <- tenemean
}
ten[is.na(ten)]<-0

###############

ten2 <- purrr::map_dbl(a,
                  ~{
                    sum(AD$rain[.:(.+1)]) / 2
                  })
ten2[is.na(ten2)]<-0

[jmcvw]

Hi AndrePaind,

I was kind of expecting this question. I wasn't sure if you were using the for loop because you were trying to specificaly learn them. Anyway, people often disparag for loops because R is designed around the idea of 'vectorization'. This is what allows x <- 1:5; x + 1 to work s it does without using an explicit loop.

sapply and friends (or their tidyverse map equivalents) are often the best alternative to loops. In this case I would suggest you look at mapply or map2 / map2_***.

The you can avoid the whole AD[i:(i+1)] thing entirely.
Eg: mapply(function(a, b) seq(a, b), a = 1:5, b = 6:10)

Sorry about not coming forward with this sooner, I wasn't holding out on you, I just thought you were on a journey...

[AndrePaind]

Thank You for the answer @nirgrahamuk but the code you suggest me, gives me a vector (ten2) full of zero.

Do you know where is the issue?

Thank You again

[nirgrahamuk]

What should I change in what I posted to create the zero's as you do ?
FAQ: How to do a minimal reproducible example ( reprex ) for beginners

[AndrePaind]

that's funny.. yep i'm on my journey to learn R (I'm coming from Matlab so i've already a bit of knowledge of programming but I miss the "grammar")

@jmcvw thanks for your answers, it's good to see unconditional help here around.
It's easier if you can talk directly with someone instead of looking million of pages on internet.

[AndrePaind]

ok so let's work with just a vector AD, which I'm gonna represent it as a list of random numbers (it's not a list of rand numbers in reality but that's ok)
If i'm running this, for example, the result is the same for every element and it should not.

a <- integer(800000)
AD <- runif (800000)
ten2 <- purrr::map_dbl(a,
                  ~{
                    sum(AD[.:(.+1)]) / 2
                  })
ten2[is.na(ten2)]<-0

[nirgrahamuk]

whats the purpose of a, it seems to have zero information content, it just coincidentally happens to be as long as the AD object ?
you should probably omit it, the role it served in my first example was the indexes into AD to manipulate.

set.seed(42)
AD <- runif (800000)
ten2 <- purrr::map_dbl(seq_along(AD),
                       ~{
                         sum(AD[.:(.+1)]) / 2
                       })
ten2[is.na(ten2)]<-0

#an alternative using slider to slide a 2-length window over AD and apply a function to it
library(slider)
ten3 <- slide_dbl(AD,
                  .after = 1
                  ,.f = mean)

note that in the slider example ten3 differs from ten2 based on the differing treatment of the final value,
which in the first way, is NA mapped to 0 but in the second way, is the mean of the single number in the last position of AD

[AndrePaind]

amazing, thanks a lot @nirgrahamuk