| Re: Urgent object question [message #28921] |
Wed, 23 January 2002 07:42  |
Ted Cary
Messages: 53
Registered: October 2001
|
Member |
|
|
"David Fanning" <david@dfanning.com> wrote in message >
> After a solid week of this we find that our ideas
> *always* look wonderful about midnight. But not all
> of them have survived the bright glare of morning. :-(
> ... We are in the process
> now of building a complicated enough system to test
> whether it works or not.
Morning is almost over, and it still seems to be a good idea. If you want a
not-so-complicated system to test, I can send you guys mine. I'm sure the
boss would be happier if *real* programmers wrote my application. Hopefully
you and your friends will publish the results on your web page, so that I
will know what I've done wrong. Who's the "we" you speak of anyway--I hope
you haven't started referring to yourself and your computer in the first
person plural...
|
|
|
|
|
|
| Re: Urgent object question [message #28925 is a reply to message #28923] |
Wed, 23 January 2002 07:06  |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Ted Cary (tedcary@yahoo.com) writes:
> Mmmm... IDL_Container? Could I just subclass an IDL_Container and my
> "planet" object widget, then fill the container with all my "satellite"
> object widgets? If each satellite had an object reference to the planet or
> (better) to this very subclass containing the IDL_Container, then all of
> them could change the planet object's fields or get its fields via
> Set/GetProperty methods. Then all that's left is to override the planet
> object's methods so that they also call methods on every object in the
> IDL_Container, and there's my two-way communication.
That is the basic idea, yes. We are in the process
now of building a complicated enough system to test
whether it works or not. For us, the bottom line is
whether this system gives us advantages in maintaining
and extending large applications.
> This also solves the problem of destroying the object widgets when their
> TLBs are destroyed by the user.
Yes, containers have many advantages and this is certainly
one of them.
> It makes sense now, but I'm also very tired.
After a solid week of this we find that our ideas
*always* look wonderful about midnight. But not all
of them have survived the bright glare of morning. :-(
Cheers,
David
--
David W. Fanning, Ph.D.
Fanning Software Consulting
Phone: 970-221-0438, E-mail: david@dfanning.com
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Toll-Free IDL Book Orders: 1-888-461-0155
|
|
|
|
| Re: Urgent object question [message #28932 is a reply to message #28925] |
Tue, 22 January 2002 21:11 |
Ted Cary
Messages: 53
Registered: October 2001
|
Member |
|
|
Mmmm... IDL_Container? Could I just subclass an IDL_Container and my
"planet" object widget, then fill the container with all my "satellite"
object widgets? If each satellite had an object reference to the planet or
(better) to this very subclass containing the IDL_Container, then all of
them could change the planet object's fields or get its fields via
Set/GetProperty methods. Then all that's left is to override the planet
object's methods so that they also call methods on every object in the
IDL_Container, and there's my two-way communication.
This also solves the problem of destroying the object widgets when their
TLBs are destroyed by the user. In my first scheme, the XManager Cleanup
procedure for each satellite would have to destroy the object part of the
object widget , a problem if the object it is trying to destroy has been
subclassed! The XManager-called cleanup routine would have to test the
object before destroying it, and then if the object was subclassed it would
have to reset all the corresponding self fields that were inherited from
just its object. Using the IDL_ Container, the satellites' self fields are
each in separate objects within the whole "planetary system" object, so they
can be destroyed along with their object and with the TLB of their Widget
interface.
It makes sense now, but I'm also very tired--probably this is what
IDL_Containers are used for all the time. Any obvious problems with this
system before I start coding it tomorrow, besides the fact that I keep
calling my objects planets and satellites? (If there are more formal
programming terms, tell me.)
Thanks.
"David Fanning" <david@dfanning.com> wrote in message
news:MPG.16b7f18d835b7a669897d9@news.frii.com...
> Ted Cary (tedcary@yahoo.com) writes:
>
>> Does anyone have experience with a problem like this? Or any
>> suggestions? Fanning?
>
> Uh, funny you should mention this. I am up to my
> eyeballs in a problem almost exactly like this one. :-(
>
> Let me get back to you when I have a little more
> time and I see the path just a little clearer than
> I do at the moment. But I've been fooling around with
> container objects and object hierarchies that communicate
> with one another in a way that is analogous to the way
> widgets communicate. But it is all still a bit unsettled....
>
> Cheers,
>
> David
> --
> David W. Fanning, Ph.D.
> Fanning Software Consulting
> Phone: 970-221-0438, E-mail: david@dfanning.com
> Coyote's Guide to IDL Programming: http://www.dfanning.com/
> Toll-Free IDL Book Orders: 1-888-461-0155
|
|
|
|
|
|
| Re: Urgent object question [message #29006 is a reply to message #28921] |
Thu, 24 January 2002 07:16 |
David Burridge
Messages: 33
Registered: January 1998
|
Member |
|
|
Hi Ted,
Ted Cary wrote in message ...
>
> "David Fanning" <david@dfanning.com> wrote in message >
>
>> After a solid week of this we find that our ideas
>> *always* look wonderful about midnight. But not all
>> of them have survived the bright glare of morning. :-(
>> ... We are in the process
>> now of building a complicated enough system to test
>> whether it works or not.
>
> Morning is almost over, and it still seems to be a good idea. If you want
a
> not-so-complicated system to test, I can send you guys mine. I'm sure the
> boss would be happier if *real* programmers wrote my application.
Hopefully
> you and your friends will publish the results on your web page, so that I
> will know what I've done wrong. Who's the "we" you speak of anyway--I hope
> you haven't started referring to yourself and your computer in the first
> person plural...
>
I just got back from that week of programming Dave mentioned, so forgive the
delay in sticking my oar in!
The way I see it, using the IDL_Container (or a subclass if you need some
specific
behaviour) has a ton of advantages. The only problem with it is 1) you need
to be
absolutely clear on the ISA vs HASA relationship and 2) you can have get/set
calls
flying everywhere - causing infinite loops and making traceback impossible!
By only passing unresolved get/set keyword requests to the parent object,
we've cut
down on the tracking problem and eliminated infinite loops. This is
relatively easy
using the keyword inheritance methods. Secondly, we've buried all the
mechanics
in a single superclass (e.g. inheriting IDL_Container and passing unresolved
keywords to parent container objects) so that the hierarchy can be created
almost
transparently by simply inheriting our top-level object. Last of all, by
accepting the
parent class as a param to the init method, our object adds ITSELF to the
container,
adding to the transparency effect.
As Dave said, this idea is still bedding down. I wrote a little test app
which proved the
concept, now we're trying to break it!-) Hope this garble is understandable
and helpful!
Dave
|
|
|
|
| Re: Urgent object question [message #29020 is a reply to message #28921] |
Wed, 23 January 2002 09:11 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Ted Cary (tedcary@yahoo.com) writes:
> Morning is almost over, and it still seems to be a good idea. If you want a
> not-so-complicated system to test, I can send you guys mine. I'm sure the
> boss would be happier if *real* programmers wrote my application.
Well, at least until the bill arrived. :-)
> Who's the "we" you speak of anyway--I hope
> you haven't started referring to yourself and your computer in the first
> person plural...
I told you I need to get a life!
Actually, I've had to engage a real programmer myself,
so I can get on with the tennis. :-)
Cheers,
David
--
David W. Fanning, Ph.D.
Fanning Software Consulting
Phone: 970-221-0438, E-mail: david@dfanning.com
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Toll-Free IDL Book Orders: 1-888-461-0155
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
