Recently, I was looking at an R function and found a if statement that looks something like this:

1
2
3

if (length(which(names(df) == "entry")) > 0) {
  # [...]
}

I guess this is common way of doing it as it matches well with a way of looking at the problem: “if I want to test the presence of a column “entry” in a data frame (df), I can check if there is more at least one name that match “entry”.

Another way, a shorter one, is to use the operator %in%:

1
2
3

if ("entry" %in% names(df)) {
  # [...]
}

I personally would have written:

1
2
3

if (sum(names(df) == "entry")) {
  # [...]
}

It is less intuitive: basically sum(names(df) == "entry") will return the number of columns names “entry” and as something that is not 0 is considered as TRUE, there is no need for > 0. I would have done so simply because I knew that sum() is quite efficient but I had never do a comparison… until today 😸! To compare the three options I wrote a small R script:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

df <- data.frame("entry" = 1, "entry2" = 1, "entry3" = 1)
nrep <- 5000000

# option 1
system.time(
for (i in seq_len(nrep)) {
  if (length(which(names(df) == "entry")) > 0) {
    1+1+1
  }
}
)

# option 2
system.time(
for (i in seq_len(nrep)) {
  if ("entry" %in% names(df))  {
    1+1+1
  }
}
)

# option 3
system.time(
for (i in seq_len(nrep)) {
  if (sum(names(df) == "entry"))  {
    1+1+1
  }
}
)

Note that system.time() is quite convenient to benchmark small pieces of code. Now, the results:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

R> # option 1
   user  system elapsed
  6.332   0.002   6.335

R> # option 2
   user  system elapsed
  5.486   0.000   5.486

R> # option 3
   user  system elapsed
  4.149   0.000   4.149

And the winner is… option 3 🏆! Interesting enough, dropping calls to functions consistently improves the efficiency but also, a smaller number of call does not mean a more efficient if statement… Not surprisingly, the efficiency of your conditional statement relies on the efficiency on the functions you call in your statement 👿!