Feature: Function to create functional sequences

[mikmart]

Inspired by this StackOverflow question, I was wondering if it might be useful to have a function to create functional sequences? AFAIK, the only way to create a fseq is to start a pipe sequence with .:

library(magrittr)

. %>%
  mean() %>% 
  format(nsmall = 3)
#> Functional sequence with the following components:
#> 
#>  1. mean(.)
#>  2. format(., nsmall = 3)
#> 
#> Use 'functions' to extract the individual functions.

For defining each step in the sequence programmatically, it might be nice to have other interfaces. Perhaps something like this to create the above?

functional_sequence(
  mean(),
  format(nsmall = 3)
)

steps <- expression(
  mean(),
  format(nsmall = 3)
)

as_functional_sequence(step)

I haven’t thought about how to actually implement these, but I was wondering if this might be something that could potentially be added to magrittr?

[ggrothendieck]

What would be nice here is a compose operator.

Consider this example. We need to use list here because otherwise by strips the class.

x <- .Date(1:3)
g <- c("a", "a", "b")
do.call("c", by(x, g, function(z) list(min(z))))

The function part can be expressed as shown below using magrittr:

library(magrittr)
do.call("c", by(x, g, . %>% min %>% list))

but it would be expressible in a more compact form if there were a compose operator. Suppose we call it $.>% and that f %.>% g is regarded to be equivalent to . %>% f %>% g. Then we could reduce the above to:

do.call("c", by(x, g, min %.>% list))