[Rd] ?Syntax wrong about `?`'s precedence ?

[Ant F]

Dear all,

`?Syntax` documents that `?` has the lowest precedence, right under `=`.

Indeed it reads:

*The following unary and binary operators are defined. They are listed in
precedence groups, from highest to lowest.  *

and ends the list with

*<- <<-* *assignment (right to left)*
*=* *assignment (right to left)*
*?* *help (unary and binary)*
I believe it to be wrong, `=` has lower precedence than `?`.

See the following example :

    `?` <- `+`
    x = 2 ? 3
    x
    #> [1] 5

We see that `2 ? 3` is evaluated first, then the result is assigned to x,
showing
higher precedence for `?`.

Compare it to the similar code using `<-` :

    x <- 2 ? 3
    #> [1] 5
    x
    #> [1] 2

Here first `x <- 2` is evaluated, then its output is added to 3, and the
result
`5` is printed. and we verify that `x` is still `2`. Showing lower
precedence
for `?` consistent with the doc.

Hadley Wickham's package `lobstr` makes it easy to compare the parse trees:

    lobstr::ast({x = 2 ? 3})
    #> o-`{`
    #> \-o-`=`
    #>   +-x
    #>   \-o-`?`
    #>     +-2
    #>     \-3

    lobstr::ast({x <- 2 ? 3})
    #> o-`{`
    #> \-o-`?`
    #>   +-o-`<-`
    #>   | +-x
    #>   | \-2
    #>   \-3

Best regards,

Antoine

	[[alternative HTML version deleted]]