Subj : Re: Calling functions without the lookup
To   : Brendan Eich <brendan@meer.net>
From : ggadwa@voyager.net (Brian Barnes)
Date : Wed Jan 15 2003 05:48 pm

Brendan Eich wrote:

> But (more super-technical detail alert) you may not be able to use 
> that API anyway, if you need to preserve the scope chain (linked by 
> the parent slot) of the object tagged in myfunc.  If you extract that 
> object's private data (a JSFunction*, call it fun), and then call that 
> function via JS_CallFunction, the API will use fun->object as the 
> callable object to invoke -- which means you'll get the scope chain 
> that the compiler saw when it created fun.
>
> That is not necessarily the right scope chain, if you are using 
> JS_CloneFunctionObject, or if the function object you got from the 
> obj.BLECH property was cloned for you automatically.  If it came from 
> a nested function declared in JS, or even a top-level function in a 
> script that you precompile in a null top-level object, or a 
> "compile-time only" top-level object used for precompiling, if the 
> script was executed against a different top-level, then the functions 
> declared in it at top-level will be automatically cloned.  Capturing 
> the value of such a function in script, and storing it in obj.BLECH, 
> could lead to a situation where you can't use JS_CallFunction from 
> native code to invoke the function with the right scope chain.
>
> In other words, JSObject => JSFunction is many to one, while 
> JSFunction => JSObject (fun->object) is a 1:1 link back to the 
> compiler-created function object.
>
> I hope this helps. 

You are the true definition of too smart for all *our* goods :)

I get it.  I shouldn't have any problems.  Basically, a script is 
compiled to an object, and that script stays static for the length of 
 run.  There's a function, always called "event", that gets called A LOT 
(sometimes, for a timer, an awful lot.)  I should be golden with what's 
above, and it saves me a lookup.  Nothing complex goes on in the 
scoping, or cloning, etc.

I've said this before, but there's a lot in spider-monkey that you don't 
get on a quick examination; the more I get into the nuts and bolts the 
better I understand how well engineered this thing was and how much 
power you can squeeze from it when you know what you are doing.  In 
short: thanks.

[>] Brian

.