using tidyverse; counting after and before change in value, within groups, generating new variables for each unique shift

Question

I am looking for a tidyverse-solution that can count occurrences of unique values of TF within groups, id in the data datatbl

Answer 1

I had a answer without data.table but it was not using dplyr. Here is my attempt using dplyr:

        #Remove the NAs 
dfr <-  df %>% filter(!is.na(TF)) %>% 
  # group by id
  group_by(id) %>% 
  # Calculate the rle on TF for each group
  do(., mrle = rle(.$TF)) %>% mutate(Total=sum(mrle$lengths)) %>%
  # Trasform the rle result in a data.frame counting the values after and before changes
  do( {
  t<- .$mrle
  #for each length generate the columns
  res <- as.data.frame(lapply(seq_along(t$lengths[-length(t$lengths)]), function(i) {

      #before change counts
      n1 <- t$lengths[i]
      #position  the counts
      if(i==1) {
        before <- 0
      } else {
        before <- sum(t$lengths[1:i-1])
      }

      #after change conts
      n2 <- t$lengths[i+1]

      if(i == (length(t$lengths)-1))
        after  <- 0
      else
        after <- .$Total - before - n1 - n2

      # assemble the column
      c(rep(NA,before),-n1:-1,1:n2, rep(NA,after))

    } ))

  colnames(res) <- paste0("PM", 1:ncol(res))
  #preserve the id
  cbind(id=.$id,res)

 })

#Join with the original data.frame
res <-  df %>% mutate(rn = row_number()) %>% filter(!is.na(TF)) %>% bind_cols(dfr) %>% right_join( df %>% mutate(rn = row_number()) ) %>% select(-rn, -id1)

#Verify
mapply(all.equal, dfa,res)
#  id   TF PM01 PM02 PM03 PM04 PM05 
#TRUE TRUE TRUE TRUE TRUE TRUE TRUE