Weird nested function behavior in IE

[Ryan Breen]

The following:

if(true)
{
outer.prototype.inner=hello_world;
function outer()
{
this.inner();
}

function hello_world()
{
alert("HI");
}

var hold_me =new outer();
}

works in IE but fails in Firefox or Rhino. In Rhino, the error is
that 'outer' is evaluated as a FUNCTION_EXPRESSION_STATEMENT rather
than a FUNCTION_STATEMENT but fn.itsClosure is null. Has anyone seen
anything like this before?

FYI, I'm using 1.5R3.

[Igor Bukanov]

Note that this code does not confirm to ECMAScript standard since
ECMAScript allows function statements only at the top level of script or
function.

Now MS JScript and SpiderMonkey/Rhino extends that allows to define
function statements anywhere were a JS statement is allowed, but the
semantics is different:

In JScript all such function statements are treated in the same way and
effectively evaluated at the start of script/function so you can refer
the functions by names right away.

In SM/Rhino such functions are defined only when the corresponding code
reaches the point of the "extended function statement" declaration which
allows to have a conditional compilation.

For example given:

if (condition) {
function test1() {
}
}

function test2() {
}

then JScript would always define test1 even if condition is false while
SM/Rhino would define it only when condition is true.

To make you example to work move hello_world()/outer to the beginning of
the if block.

Regards, Igor

[Ryan Breen]

On Tue, 02 Nov 2004 23:42:27 +0100, Igor Bukanov wrote:
> To make you example to work move hello_world()/outer to the beginning of
> the if block.

Igor,

Thanks for the feedback. Unfortunately, that's not an option in this case
-- I'm parsing existing, IE-specific JScript. Is my only option
deep-Rhino-hacking to treat all functions as FUNCTION_STATEMENTS rather
than FUNCTION_EXPRESSION_STATEMENTS?

Ryan

[Igor Bukanov]

Ryan Breen wrote:
> Unfortunately, that's not an option in this case
> -- I'm parsing existing, IE-specific JScript. Is my only option
> deep-Rhino-hacking to treat all functions as FUNCTION_STATEMENTS rather
> than FUNCTION_EXPRESSION_STATEMENTS?

Yes, that should do the trick.

Regards, Igor