CODEWITHSUNDEEP
rpipetidyversemagrittr
kiero
kiero

Reputation: 121

Strange(?) behaviour of `summary()` inside a pipe

Me and a couple of colleagues are wondering why

N <- rnorm(16) %>% matrix(., ncol = 4) %>% `colnames<-`(letters[1:4]) %T>% summary()

or

N <- rnorm(16) %>% matrix(., ncol = 4) %>% `colnames<-`(letters[1:4]) %T>% summary() %>% `+`(., 0)

do not work (summary is not printed), while

N <- rnorm(16) %>% matrix(., ncol = 4) %>% `colnames<-`(letters[1:4]) %T>% {print(summary(.))}

does?

Upvotes: 0

Views: 150

Answers (1)

Gregor Thomas
Gregor Thomas

Reputation: 145775

There's no reason for summary to print its output any more than rnorm or matrix or any of the other functions in the pipeline. Printing is generally suppressed when there is assignment <-.

N <- rnorm(16) %>% matrix(., ncol = 4) %>% `colnames<-`(letters[1:4]) %T>% summary()
# no printing

N <- 5
# no printing

M <- rnorm(16) %>% matrix(., ncol = 4) %>% `colnames<-`(letters[1:4])
# no printing

M %>% summary
summary(M)
# prints

sM <- M %>% summary
sM <- summary(M)
# no printing
# assignment prevents implicit printing.

An explicit print call creates a side-effect (printing), which is what you want to happen.

The tee pipe %T>% doesn't create any side-effects (like printing), it just returns the LHS so that if you %T>% into a function that does create side-effects but does not return it's input (like plot()), you can get those side effects while still piping the input to another step.

Note that print does return its argument (invisibly), so you don't actually need the tee pipe %T>% with print.

Upvotes: 9

Related Questions