| Re: Leaking objects... [message #29928] |
Wed, 27 March 2002 09:15 |
btupper
Messages: 55
Registered: January 2002
|
Member |
|
|
On Wed, 27 Mar 2002 09:29:39 -0700, David Fanning <david@dfanning.com>
wrote:
>
> You may *think* you want Obj_New to block the creation
> of a new object, but I can assure you, you *don't*. :-)
> Having only one of any particular kind of object would
> be just a tad limiting, don't you think?
>
> IDL is an inherently dangerous language. You really don't
> have any choice but to make an attempt to educate the
> people who use it. If we didn't believe this, why else
> would we all be hanging out here on the newsgroup?
Hi,
Since I am from the brute-force programming school, I immediately
thought 'What a great idea!' Having fiddled with this for a few
minutes, now I can see David's point. But a lot can be learned from
trying... now I know what that CAST keyword does.
Here's a hack that has half the behavior you describe.
If you set the destroy_previous keyword, the object will kill destroy
any existing objects of the TEST class that are not the new object.
However, if you do not set the /DESTROY_PREVIOUS keyword a new object
will *not* be created, but you will lose your reference, since the
failed object initialization returns a null object.
Ben
*****make an object
IDL> a = obj_new('test', findgen(20))
% Compiled module: TEST__DEFINE.
IDL> help, /heap
Heap Variables:
# Pointer: 1
# Object : 1
<ObjHeapVar23> STRUCT = -> TEST Array[1]
<PtrHeapVar24> FLOAT = Array[20]
IDL> help, a
A OBJREF = <ObjHeapVar23(TEST)>
*****try to make a new one of the same class
IDL> b = obj_new('test', findgen(30))
% TEST::INIT: An instance of this class already exists
IDL> help, a,b
A OBJREF = <ObjHeapVar23(TEST)>
B OBJREF = <NullObject>
IDL> help, /heap
Heap Variables:
# Pointer: 1
# Object : 1
<ObjHeapVar23> STRUCT = -> TEST Array[1]
<PtrHeapVar24> FLOAT = Array[20]
*****make a new object and destroy existing objects of the same class
*****note that a has been cleaned up
IDL> b = obj_new('test', findgen(30), /destroy_previous)
IDL> help, a,b
A OBJREF = <ObjHeapVar23>
B OBJREF = <ObjHeapVar26(TEST)>
IDL> help, /heap
Heap Variables:
# Pointer: 1
# Object : 1
<ObjHeapVar26> STRUCT = -> TEST Array[1]
<PtrHeapVar27> FLOAT = Array[30]
****** now try to redefine b (an existing object of class TEST
****** note that even though a new object is not created, the
****** null object defintion is returned to variable b
****** so you are really no further ahead
IDL> b = obj_new('test', findgen(40))
% TEST::INIT: An instance of this class already exists
IDL> help, b
B OBJREF = <NullObject>
IDL> help, /heap
Heap Variables:
# Pointer: 1
# Object : 1
<ObjHeapVar26> STRUCT = -> TEST Array[1]
<PtrHeapVar27> FLOAT = Array[30]
START HERE
pro test::cleanup
ptr_free, self.data
end
function test::init, data, destroy_previous = destroy_previous
objs = Obj_Valid(cast =1, count = count)
For i = 0L, Count-1 Do Begin
If Obj_Valid(objs[i]) EQ 1 Then Begin
If OBJ_CLASS(objs[i]) EQ 'TEST' AND objs[i] NE self
Then Begin
If KeyWord_Set(Destroy_Previous) Then Begin
Obj_Destroy, objs[i]
EndIf Else Begin
message, 'An instance of this class
already exists', /info
return, 0
EndElse
EndIf
EndIf
EndFor
if n_params() ne 1 then begin
message, 'Expecting 1 parameter', /info
return, 0
endif
self.data = ptr_new(data)
return, 1
end
pro test__define
j = { test, data: ptr_new() }
end
END HERE
|
|
|
|
| Re: Leaking objects... [message #29930 is a reply to message #29928] |
Wed, 27 March 2002 08:29  |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Randall Skelton (rhskelto@atm.ox.ac.uk) writes:
> I gave some IDL object code to someone who has no idea what object
> programming is. Despite giving her a quick lesson on object life-cycles,
> she did the inevitable and directly overwrote an object, creating dangling
> pointers.
>
> My question is, how do I prevent this from happening in my code? This
> behavior seems a little fragile to me. Ideally, I would like obj_new to
> either block the creation of a new object or cleanup pre-existing objects.
> What I don't want to do is try and educate the user again... sorry mom.
You may *think* you want Obj_New to block the creation
of a new object, but I can assure you, you *don't*. :-)
Having only one of any particular kind of object would
be just a tad limiting, don't you think?
IDL is an inherently dangerous language. You really don't
have any choice but to make an attempt to educate the
people who use it. If we didn't believe this, why else
would we all be hanging out here on the newsgroup?
Cheers,
David
P.S. You might suggest to your user that she become
more familiar with HEAP_GC. That might help. :-)
--
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
