| Re: removing objects from a container [message #48997] |
Sat, 10 June 2006 21:43  |
Robbie
Messages: 165
Registered: February 2006
|
Senior Member |
|
|
Going off topic again, Karl, you've incidently raised a very good
point. Actually checking if the PARENT is null is probably better than
version checking IDL. There are also other reasons why the PARENT might
be null and I need to account for this. My code will be more stable if
I avoid version checking.
> So, except for the communications
> problem, you should be in better spot by getting the fix you requested.
I am very grateful that the bug was fixed. I only checked the release
notes after I had isolated the bug, so there was certainly no harm
done. Then I realised that writing a bug report about this issue was
really silly and you guys have better and more important things to
worry about.
Cheers,
Robbie
|
|
|
|
| Re: removing objects from a container [message #48998 is a reply to message #48997] |
Sat, 10 June 2006 19:12  |
Karl[1]
Messages: 79
Registered: October 2005
|
Member |
|
|
Robbie wrote:
> Indeed, this problem had caused me much grief and frustration in the
> past. I posted a bug report to ITT that a consequence of PARENT being
> null is that IDLgrROIGroup::GetCTM() will not return the affine
> transform with respect to the root scene. This meant that my code which
> deals with converting mouse clicks to atom coordinates has a special
> case for IDLgrROIGroup (and IDLgrColorbar). The same code must now also
> detect IDL 6.3 or above to account for the new behaviour.
Yes, you were the one who reported the original bug, I think, and we
fixed it for IDL 6.3.
This is a slight change in subject from the Remove method discussion
because you're bringing up the issue of the PARENT property.
It is pretty hard to fix a bug while maintaining
backwards-compatibility, because we're purposely fixing the previous
bad behavior by changing it.
I don't know how your special case code works, but one option may be to
simply leave it in place and it may work the same with the 6.3 bug fix
in place. Other options include checking the PARENT property for NULL
and taking different directions based on that, or implementing an IDL
version check.
Bugs happen and we do our best to fix them. While we'd like IDL
applications to be free of version-specific paths and workarounds, the
reality is otherwise, but we strive to keep it to a minimum.
> So, yes! I'm definately getting different behaviour in IDL 6.3. The
> curious thing is that the fix was applied *after* IDL 6.3 Beta and
> isn't documented anywhere I could see. Am I playing with fire here?
I don't think so. We do fix bugs after Beta. Otherwise, there
wouldn't be much point to a Beta program, right? :-) I am, however,
not certain why the fix didn't make it into the release notes or why it
didn't otherwise get communicated. All I can do is apologize for that
and point out that you'd have to take the same steps to adapt your
application to the fix anyway, even if there had been an entry in the
release notes about the fix. So, except for the communications
problem, you should be in better spot by getting the fix you requested.
Karl
(ittvis)
|
|
|
|
|
|
| Re: removing objects from a container [message #49004 is a reply to message #49000] |
Fri, 09 June 2006 16:38 |
Karl Schultz
Messages: 341
Registered: October 1999
|
Senior Member |
|
|
On Fri, 09 Jun 2006 14:08:01 -0400, Ben Tupper wrote:
> Hello,
>
> David - please look out the window at your roses for a few minutes...
>
> I can't quite figure the REMOVE method out. IDLgrROIgroup inherits from
> (eventually) IDL_CONTAINER so the REMOVE method is supposed to work. I
> don't see in the docs that IDLgrROIgroup overrides the method - but
> something is different.
>
>
> IDL> group = OBJ_NEW('IDLgrROIgroup')
> IDL> roi = OBJ_NEW('IDLgrROI')
> IDL> group->Add, roi
> IDL> group->REMOVE, /ALL
> % IDLGRROIGROUP::REMOVE: Incorrect number of arguments.
> % Execution halted at: $MAIN$
> IDL> group->IDL_CONTAINER::REMOVE,/ALL
>
>
> Is this a bug or just a thorn?
>
It looks like a new bug was introduced here in 6.3 while trying to fix
another problem.
The graphic objects have a PARENT property that is managed by the
IDLgrContainer class as graphic objects are added or removed to/from
containers like grModel. IDLgrROIGroup can't subclass from
IDLgrContainer because IDLgrROIGroup inherits from IDLanROIGroup, which
inherits from IDL_Container (IDLanROIGroup does not store displayable
graphic objects) So, IDLgrROIGroup inherits from IDL_Container for its
container functionality, and IDL_Container doesn't manage a PARENT
property.
So, prior to 6.3, the PARENT property of an IDLgrROI was never getting
updated as it was added or removed from an IDLgrROIGroup.
We fixed this bug in 6.3 by updating the PARENT property in the
already-overridden Add method and adding a new override for the Remove
method that would clear the PARENT property when an IDLgrROI object was
removed.
Unfortunately, the form IDLgrROIGroup->Remove, /ALL was not implemented
correctly in the new Remove method. And it probably was not documented
because it was supposed to work exactly like the superclass method.
I'll open a change request and fix this for next release.
The only workaround I can think of for now is:
group->Remove, group->Get(/ALL)
This would work correctly with or without the bug.
If you do
group->IDL_CONTAINER::REMOVE,/ALL
then the PARENT property of all the ROI's that were in the group will be
left referring to the group object. If that doesn't bother you, then this
would be another way to do it, but I don't think it is a good idea, as it
leaves potential dangling object references.
Karl
|
|
|
|
| Re: removing objects from a container [message #49008 is a reply to message #49004] |
Fri, 09 June 2006 12:07 |
Michael Galloy
Messages: 1114
Registered: April 2006
|
Senior Member |
|
|
David Fanning wrote:
> Ben Tupper writes:
>
>> David - please look out the window at your roses for a few minutes...
>
> OK, I even gave the dog a good scratch while I was at it. I *do* feel
> better. :-)
>
>> I can't quite figure the REMOVE method out. IDLgrROIgroup inherits from
>> (eventually) IDL_CONTAINER so the REMOVE method is supposed to work. I
>> don't see in the docs that IDLgrROIgroup overrides the method - but
>> something is different.
>>
>>
>> IDL> group = OBJ_NEW('IDLgrROIgroup')
>> IDL> roi = OBJ_NEW('IDLgrROI')
>> IDL> group->Add, roi
>> IDL> group->REMOVE, /ALL
>> % IDLGRROIGROUP::REMOVE: Incorrect number of arguments.
>> % Execution halted at: $MAIN$
>> IDL> group->IDL_CONTAINER::REMOVE,/ALL
>>
>>
>> Is this a bug or just a thorn?
>
> Well, bad news. Here is my IDL session, with *no* error message!
>
> IDL> group = OBJ_NEW('IDLgrROIgroup')
> IDL> roi = OBJ_NEW('IDLgrROI')
> IDL> group->Add, roi
> IDL> group->REMOVE, /ALL
>
> IDL 6.2 on Windows. (I'm not getting a new version of IDL until
> there is a good reason to. :-)
>
> Cheers,
>
> David
I get IDL 6.2 on Mac working and IDL 6.3 on Mac failing with the same
error message.
I don't see anything in the docs about a change.
Mike
--
www.michaelgalloy.com
|
|
|
|
| Re: removing objects from a container [message #49010 is a reply to message #49008] |
Fri, 09 June 2006 11:19 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Ben Tupper writes:
> David - please look out the window at your roses for a few minutes...
OK, I even gave the dog a good scratch while I was at it. I *do* feel
better. :-)
> I can't quite figure the REMOVE method out. IDLgrROIgroup inherits from
> (eventually) IDL_CONTAINER so the REMOVE method is supposed to work. I
> don't see in the docs that IDLgrROIgroup overrides the method - but
> something is different.
>
>
> IDL> group = OBJ_NEW('IDLgrROIgroup')
> IDL> roi = OBJ_NEW('IDLgrROI')
> IDL> group->Add, roi
> IDL> group->REMOVE, /ALL
> % IDLGRROIGROUP::REMOVE: Incorrect number of arguments.
> % Execution halted at: $MAIN$
> IDL> group->IDL_CONTAINER::REMOVE,/ALL
>
>
> Is this a bug or just a thorn?
Well, bad news. Here is my IDL session, with *no* error message!
IDL> group = OBJ_NEW('IDLgrROIgroup')
IDL> roi = OBJ_NEW('IDLgrROI')
IDL> group->Add, roi
IDL> group->REMOVE, /ALL
IDL 6.2 on Windows. (I'm not getting a new version of IDL until
there is a good reason to. :-)
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
|
|
|
|
| Re: removing objects from a container [message #49087 is a reply to message #49004] |
Mon, 12 June 2006 07:36 |
btt
Messages: 345
Registered: December 2000
|
Senior Member |
|
|
Karl Schultz wrote:
>
> I'll open a change request and fix this for next release.
>
> The only workaround I can think of for now is:
>
> group->Remove, group->Get(/ALL)
>
> This would work correctly with or without the bug.
>
> If you do
>
> group->IDL_CONTAINER::REMOVE,/ALL
>
> then the PARENT property of all the ROI's that were in the group will be
> left referring to the group object. If that doesn't bother you, then this
> would be another way to do it, but I don't think it is a good idea, as it
> leaves potential dangling object references.
>
Thanks, Karl.
I shall heed your advice and modify my code accordingly.
Cheers,
Ben
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
