concurrency issues

**Chris Smith** · Jul 20 '05, 09:23 AM

Re: concurrency issues

Andrew Urquhart wrote:[color=blue][color=green]
> > The Problem:
> >
> > How does Frame B find out when Frame A has finished creating the object?
> > If Frame A hasn't finished creating the object yet, I want Frame B to
> > wait, and invoke the method as soon as the object has been created. I
> > can find out IF the object has been created (test the variable for
> > undefined or null), but if it hasn't been created yet, I can't find a way
> > to wait and send the notification when it has.[/color]
>
> Hi Chris, good - a nice clear description.
>
> In frame B:
> In a timed loop within Frame B, poll frame A every 200 milliseconds or so
> to see if the object exists (actually poll each stage down the DOM tree to
> catch if the frame exists etc, before directly testing for the object).
> Only when you find the object do you leave the polling loop and invoke the
> rest of your code. Note: You might want to put a limit on how long you poll
> before you give up and err. You should also check that the method is
> available on your object, rather than just assuming that once the object
> exists that the method will as well. Something similar I think to some old
> code I once wrote: http://www.andrewu.co.uk/clj/checkframesloaded/[/color]

Does the code at the end of this message look like what you were
suggesting? Do you see any problems with it?

Incidentally, you wrote:[color=blue]
> Only when you find the object do you leave the polling loop and invoke the
> rest of your code.[/color]

I assume by "the rest of your code", you mean the method I need to
invoke on the other frame? I certainly want the main frame to be
operational before the other frame finishes loading.
[color=blue]
> You should also check that the method is
> available on your object, rather than just assuming that once the object
> exists that the method will as well.[/color]

Okay, I've done so... but is there any reason this would ever happen?
I'm assigning the methods to the object in the constructor, and I would
think that the constructor would have to completely finish before the
reference returned from the object creation expression. I'm not very
familiar with JavaScript, and I'll do it the careful way until I know
for sure...

Anyhow, here's the code...

<script type="text/javascript">
var topDone = false;
var bottomDone = false;
var intervalKey = window.setInter val(function()
{
if (!topDone && parent && parent.topFrame
&& parent.topFrame .menu && parent.topFrame .menu.showPage)
{
parent.topFrame .menu.showPage( '<%= pageName %>');
topDone = true;
}

if (!bottomDone && parent && parent.bottomFr ame
&& parent.bottomFr ame.menu && parent.bottomFr ame.menu.showPa ge)
{
parent.bottomFr ame.menu.showPa ge('<%= pageName %>');
bottomDone = true;
}

if (topDone && bottomDone)
{
window.clearInt erval(intervalK ey);
}
}, 200);
</script>

I'm a bit bothered by polling, even with the delay... but I guess it
can't be helped, then.

--

https://www.designacourse.com

The Easiest Way to Train Anyone... Anywhere.

Chris Smith - Lead Software Developer/Technical Trainer
MindIQ Corporation

**Andrew Urquhart** · Jul 20 '05, 09:23 AM

Re: concurrency issues

On Mon, 28 Jul 2003 14:53:45 -0600, Chris Smith <cdsmith@twu.ne t> wrote:
[color=blue][color=green]
>> Something similar I think to some old code I once wrote:
>> http://www.andrewu.co.uk/clj/checkframesloaded/[/color]
>
> Does the code at the end of this message look like what you were
> suggesting?[/color]

Yes, I think so.
[color=blue]
> Do you see any problems with it?[/color]

I see you're working backwards up the DOM tree from your current context,
which is probably safer than what I would have done - it means that your
script should still work if your site is viewed in someone else's frame!
[color=blue]
> Incidentally, you wrote:[color=green]
>> Only when you find the object do you leave the polling loop and invoke
>> the rest of your code.[/color]
>
> I assume by "the rest of your code", you mean the method I need to invoke
> on the other frame? I certainly want the main frame to be operational
> before the other frame finishes loading.[/color]

I was assuming that the call itself to the method living in frame A would
be just a part of additional script in frame B. If all frame B wants to do
is invoke the object method in frame A once it's available, then yes. Not
sure what the implications are of your last sentence there as I don't know
what frame you're referencing when you say "main" frame.
[color=blue][color=green]
>> You should also check that the method is available on your object,
>> rather than just assuming that once the object exists that the method
>> will as well.[/color]
>
> Okay, I've done so... but is there any reason this would ever happen?[/color]

I didn't know how you were setting your object methods, so assumed the
worst case scenario. You could have created the object and then added
methods to it later on, in which case your script had a very small chance
of failing since the object could exist before the method did! Although it
could be considered pedantic, personally I consider it just good practice.

<trim>
[color=blue]
> I'm a bit bothered by polling, even with the delay... but I guess it
> can't be helped, then.[/color]

OK, but given the nature of the medium I don't think you can do much else
(at least not that is cross-browser, although you could always find a non-
framed solution!) Reduce the polling delay if needs be (but not too small,
the granularity on some platform combos can be of the order of 100ms if I
remember rightly). Also bear in mind that if this is running in an open web
environment I'd be surprised if anyone noticed an extra 100-200ms when it
might take ten times longer to download the page, and don't forget the
average additional wait as a result of this script is half this value!

Cheers,
--
Andrew Urquhart
- http://www.andrewu.co.uk/clj.asp
- FAQ for comp.lang.javas cript by Jim Ley at http://jibbering.com/faq
- Archive at http://groups.google.com/groups?grou...ang.javascript

concurrency issues

concurrency issues

Comment

Comment