Short circuit evaluation with an assignment statement

General Usage
1

There is an example in the (increasingly more excellent) manual under short circuit evaluation that shows an assignment statement as the follow-on statement:

julia> true && (x = (1, 2, 3))
(1, 2, 3)

julia> false && (x = (1, 2, 3))
false

The example is clear enough, but the manual doesn’t say why an assignment statement is different than a statement with side effects, say a println(“something”). Just curious about why…

1 Like
2

Not sure what the exact question is but one “difference” is that you need the paranthesis on the right hand side when making an assignment.

3

The question is WHY are the parens needed for the statement on the right side when it is an assignment statement as opposed to, say, a function call.

4

Because of presedence. Otherwise it is

(false && x) = (1,2,3)

1 Like
5

Interesting that = doesn’t have higher precedence than a boolean operator. But, it is what it is and has been since the beginning.

Done.

6

It makes sense for assignment to have lower presedence, consider a = b==c, you don’t want that to be read as (a=b) == c

2 Likes
7

Or even more pertinent to this question, you want

x = y && z

to parse as x = (y && z) not as (x = y) && z. Therefore, = must have lower precedence than && and you also have

x && y = z

parsing as (x && y) = z rather than x && (y = z). If && and = had adjacent precedence levels, you could just put them on the same level and make them left-associative, but they are separated by several precedence levels so that’s not feasible. It might just be possible to break the usual precedence parsing model and give = different precedence on the LHS here than on the RHS, but that’s getting into pretty dicey parsing territory.

1 Like
8

I think parens are fine to be explicit about the order of operation intended: as in nearly every expression. I was lulled into thinking of short circuit evaluation as a way to “run” commands, not as an expression in its own right. As
the logic table in the documentation makes clear, it is a Boolean expression. It’s just a bit of a trick that we can employ it as a device to run commands as a shortcut to a slightly longer if block. I was merely surprised by the obvious. No serious or
valid recommendation to change anything here.

Thanks, all.

1 Like
9

The presedence and parsing of = has been used constructively by many domain-specific languages. Below is an example from Modia that makes use of the fact that julia happily parses any valid expressio on the lhs of = (If I’m not mistaken, one of the many reasons Julia was chosen as the host language of Modia)

@model FirstOrder begin
     x = Variable(start=1)   # start means x(0)
     T = Parameter(0.5)      # Time constant
     u = 2.0                 # Same as Parameter(2.0)
  @equations begin
     T*der(x) + x = u        # der() means time derivative
     end
  end;
1 Like