Behavior of `eval` in non strict mode.

# Benjamin (Inglor) Gruenbaum (8 years ago)

I've recently run into this question in Stack Overflow:

stackoverflow.com/q/21008329/1348195


function f() {
    f = eval("" + f);
    console.log("Inside a call to f(), f is: \n%s", f);}

f();

console.log("After a call to f(), f is: \n%s", f);

What should the output of the following be?

I expected undefined on both but that's because I'm used to strict mode. IE/Chrome treat this differently from Firefox and to be honest when I checked the spec it boiled down to which context is affected here.

In IE/Chrome the eval is creating f inside the context of f acting like a function declaration inside. In Firefox it's acting like it's running in the global context.

Which is correct? I've tried to follow 10.4.2 (or 18.2.1 in the ES6 draft which is nice) but I still couldn't figure out what "if there is no calling context means".

# Andrea Giammarchi (8 years ago)

looks rather an eval gotcha but I think Firefox is correct anyway. try f = eval("(" + f + ")"); instead and it should produce what you expect (I guess)

# Benjamin (Inglor) Gruenbaum (8 years ago)

Thanks for the reply.

I'd actually expect undefined because function declarations does not return anything. Converting it to a function expression kind of misses the point since those are well... expressions :)

I've tried looking in all the relevant places in the spec but still couldn't unambiguously figure out which browser is 'correct'.

# Andrea Giammarchi (8 years ago)

I think eval returns whatever it evaluates ... i.e.

var x = eval('123');

x will be 123 since it's returned. Accordingly, if you assign a function, this should be returned and become automatically an expression.

The inconsistency exists using explicitly parenthesis but I don't remember specs saying that eval should not return the evaluated content in case of function declaration.

Hard to tell which one is correct, the example is odd anyway.

# André Bargull (8 years ago)

Thanks for the reply.

I'd actually expect undefined because function declarations does not return anything. Converting it to a function expression kind of misses the point since those are well... expressions :)

I've tried looking in all the relevant places in the spec but still couldn't unambiguously figure out which browser is 'correct'.

There are a few edge cases in reference resolution which are not correctly implemented in most browsers. Your example is basically the same as test case 2 from ecmascript#1751. The relevant section in the specification is "12.13.4 Runtime Semantics: Evaluation": The left hand side of an assignment is always evaluated before the right hand side. This includes resolving and remembering the reference information for an identifier reference. In this case the identifier reference resolves to a binding on the global object, so assignment must be performed on the global, even if a (direct) eval expression introduces new bindings with the same name in the current scope.

# Brendan Eich (8 years ago)

Has anyone filed bugs against V8 and Chakra?

# Benjamin (Inglor) Gruenbaum (8 years ago)

Thanks, this clarifies things. I'll update the answer on SO to reflect the findings.

# Andrea Giammarchi (8 years ago)

I've learned it the hard way ... when in doubt, see what Firefox does ... usually that's the meant standard behavior.

I really wish JavaScript was a Test Driven developing programming language ... the amount of fragmentation for every single little thing apparently never tested against meant specs is often too damn high!

# Brendan Eich (8 years ago)

Things to complain about!

JS interop is far better than other languages with multiple implementations. Never mind complex APIs such as the DOM.