[Rd] [External] Re: New pipe operator

[Duncan Murdoch]

On 07/12/2020 12:03 p.m., Gregory Warnes wrote:
> My vote is for the consistency of function calls always having parentheses,
> including in pipes.  Making them optional only saves two keystrokes, but
> will add yet another inconsistency to confuse or trip folks up.
>
> As for the new anonymous function syntax, I would prefer something more
> human friendly, perhaps provide “fun” as a shortcut for “function”,
> enabling:
> 
> DF <- "myfile.csv" %>%
>      readLines() |>
>      fun(x) gsub(r'{(c\(.*?\)|integer\(0\))}', r'{"\1"}', x) |>
>      fun(x) read.csv(text = x)|>
>     mutate(
>          across(2:3,
>                fun(col) lapply(col,
>                     fun(x) eval(parse(text = x))
>                )
>          )
>     )
> 
> which seems much easier to read and understand, at the cost of only a few
> extra characters.

But you didn't "always" include parentheses, you skipped them on the 
calls to the anonymous functions.  I think that's the one place I'd make 
the exception, so maybe we agree:  parens are almost always needed, with 
the sole exception being anonymous functions.

As to using "fun", I think that's a bad idea.  I haven't checked, but I 
wouldn't be too surprised if "fun" has been used thousands of times in 
CRAN packages as the name of a function.  So

  x |> fun(y)

would mean "fun(x, y)", whereas

  x |> fun(y) y+1

would mean (function(y) y+1)(x).

Duncan Murdoch