Inconsistency in Results Using sum(. != 3) within Pipeline Operations (%>%) dplyr version 1.1.4

[ZHBHSMILE]

I encountered an inconsistency in the results obtained from different approaches while working with R code. Specifically, when using the sum(. != 3) expression within a pipeline operation (%>%), the outcome varied from expectations, leading to discrepancies compared to direct computations.

  data <- data.frame(
    GO.BiologicalProcess = c("-", "-", "A", "B"),
    GO.CellularComponent = c("-", "C", "-", "D"),
    GO.MolecularFunction = c("-", "-", "-", "E")
  )
  
  # Calculate the number of occurrences of "-" in each row
  go_num <- rowSums(data[, c("GO.BiologicalProcess", "GO.CellularComponent", "GO.MolecularFunction")] == "-")
  
  # Calculate the sum of occurrences where the count is not equal to 3
  go_bg_num <- sum(go_num != 3)
  
  # Using pipe operator and sum(. != 3)
  error_sum <- rowSums(data[, c("GO.BiologicalProcess", "GO.CellularComponent", "GO.MolecularFunction")] == "-") %>% sum(. != 3)

[SpatLyu]

Calculate it like this:

library(magrittr)

data <- data.frame(
  GO.BiologicalProcess = c("-", "-", "A", "B"),
  GO.CellularComponent = c("-", "C", "-", "D"),
  GO.MolecularFunction = c("-", "-", "-", "E")
)

{data[, c("GO.BiologicalProcess", "GO.CellularComponent", "GO.MolecularFunction")] == "-"} %>% 
  rowSums() %>% 
  {. != 3} %>% 
  sum()
#> [1] 3

^{Created on 2024-06-10 with reprex v2.1.0}

[philibe]

initial datas

  foo<-{data[, c("GO.BiologicalProcess", "GO.CellularComponent", "GO.MolecularFunction")] == "-"} %>% rowSums() 
  foo
  # [1] 3 2 2 0

{. != 3} make boolean :

  foo2 <-  foo  %>%    {. != 3}
  foo2
  # [1] FALSE  TRUE  TRUE  TRUE

Therefore sum(foo2) is the number of TRUE = 3

At the opposite

  foo %>%  {. != 3} %>% sum(.=TRUE)
  # [1] 4

And in basic R or basic R with %>%

  sum(foo[foo!=3] )
  # [1] 4
  foo[foo!=3] %>% sum(.)
  # [1] 4

or piped function or not piped function:

  testsum <- function (x) if_else(x!=3,x,0)
  foo %>% testsum(.) %>% sum(.)
  # [1] 4
  sum(testsum(foo))
  # [1] 4

[DavisVaughan]

Read ?`%>%` closely, particularly Using the dot for secondary purposes and you will see that this is actually known and expected behavior! The LHS is being placed as the first argument of sum(), and as . in the second argument to sum() (the . != 3 expression).

You can use braces to avoid this

rowSums(data[, c("GO.BiologicalProcess", "GO.CellularComponent", "GO.MolecularFunction")] == "-") %>% {
  sum(. != 3)
}

but I think avoiding the pipe entirely is cleaner here

sums <- rowSums(data[, c("GO.BiologicalProcess", "GO.CellularComponent", "GO.MolecularFunction")] == "-")
sum <- sum(sums != 3)

[philibe]

Thanks for the explanation. I've looked for why c(3, 2, 2, 0) %>% sum(.!=3)=10, which seemed very weird. :)