Assigning values from if-else blocks (how does it work?)

Answer 1

Let me focus on

What is bugging me is that I can't work out why this code works—I was amazed that it doesn't just throw an error.

Why do you think so? What happens is that we have

1. y being assigned an expression
2. that expression being an if if() ...
3. leading to either a TRUE or FALSE in the test
4. leading to either one of the two branches being entered
5. leading to the respective code being evaluated
6. leading to its final value being the value of the right-hand-side
7. leading to this value being assigned to y

Seems logical to me.

Answer 2

It's not really very deep - many languages use the "val = x ? y : z" construct for this. In R, that's just folded into an if/else construct, so you write "val = if(x) y else z" instead.

Answer 3

Control structures such as if(....) ... else ... are equivalent to function calls with the relevant (lazy-evaluated) subexpressions as the argument:

`if`(TRUE, "yes", "no")

Basically, if is a special primitive function:

R> sapply(ls("package:base", all=TRUE), function(x) is.primitive(get(x)))["if"]
  if 
TRUE 
R> sapply(ls("package:base", all=TRUE), function(x) typeof(get(x)))["if"]
       if 
"special"

The relevant section (Section 3.2.1) in the R Language Definition manual says:

Because if/else statements are the same as other statements you can assign the value of them. The two examples below are equivalent.

R> if( any(x <= 0) ) y <- log(1+x) else y <- log(x)
R> y <- if( any(x <= 0) ) log(1+x) else log(x)

Answer 4

About second part of your question - {} is used to group expressions, it works diffrent than in a function definition. As you can read in help("Paren"):

For ‘{’, the result of the last expression evaluated.

But all expressions are evaluated in the current environment:

y <- if (TRUE) {
    print(paste("Current frame:",sys.nframe()))
    y <- 3
    z <- 5
    4
}
# [1] "Current frame: 0"
# result:
y # [1] 4
z # [1] 5

# compare to use function:
v <- if(TRUE) (function(){      
    print(paste("Current frame:",sys.nframe()))
    v <- 3
    w <- 5
    4
})()
# [1] "Current frame: 1"
v # [1] 4
w # Error: object 'w' not found

Conclusion is blocks don't behave like functions.

edit: If you want use blocks as function you could use local:

a <- if (TRUE) local({
    print(paste("Current frame:",sys.nframe()))
    a <- 3
    b <- 5
    4
})
# [1] "Current frame: 6"
a # [1] 4
b # Error: object 'b' not found

Then you could use return too:

a <- if (TRUE) local({
    print(paste("Current frame:",sys.nframe()))
    a <- 3
    b <- 5
    return(7)
    4
})
# [1] "Current frame: 6"
a # [1] 7
b # Error: object 'b' not found