| Re: Memory Cleanup-- Messy structure [message #71382] |
Thu, 17 June 2010 09:43  |
penteado
Messages: 866
Registered: February 2018
|
Senior Member
Administrator |
|
|
On Jun 17, 12:45 pm, Craig Markwardt <craig.markwa...@gmail.com>
wrote:
> On Jun 16, 1:50 pm, Ed Hyer <ejh...@gmail.com> wrote:
>
>> Since I'm not a real programmer, I still consider pointers to be kinda
>> voodoo. Can someone patiently explain how I manually clean up the
>> memory in this case?
>
> I still think RSI/Kodak/ITT dropped the ball on the design of this.
> IDL is a high level language, not a low level language - "4th
> generation" right? IDL should be cleaning up messes like this, not
> us.
I think it is proper for the times when IDL originated, when it was
too expensive to do automatic garbage collection. IDL 8 has it (it can
be switched off), though one could argue that it should not have taken
so long.
|
|
|
|
|
|
|
|
| Re: Memory Cleanup-- Messy structure [message #71400 is a reply to message #71389] |
Wed, 16 June 2010 14:53  |
MarioIncandenza
Messages: 231
Registered: February 2005
|
Senior Member |
|
|
CR,
Thanks much. This turned out, of course, to be PEBKAC. UNDEFINE does
in fact do its job, if you know how to use it. However, I did find out
this one nugget: You can see that from the way the array is created
and populated, it looks like it will have nested pointers. However, if
you do it the other way:
IDL> STRUCT1=CREATE_STRUCT("PTR1",PTR_NEW(/ALLOC));
IDL> STRUCTALL=REPLICATE(STRUCT1,N)
IDL> FOR I=0,N-1 DO *(STRUCTALL[I].PTR1)=ALLDATA[*,I]
You will find that there is only one pointer, which after that
operation points to the last data tested = ALLDATA[*,N-1].
Later on I will post the fancy new routine that led me to have to
solve this problem!
Thanks,
--Edward H.
|
|
|
|
| Re: Memory Cleanup-- Messy structure [message #71402 is a reply to message #71400] |
Wed, 16 June 2010 14:16 |
rogass
Messages: 200
Registered: April 2008
|
Senior Member |
|
|
On 16 Jun., 19:50, Ed Hyer <ejh...@gmail.com> wrote:
> OK, here's the sequence:
>
> IDL> STRUCT1=CREATE_STRUCT("PTR1",PTR_NEW());
> IDL> STRUCTALL=REPLICATE(STRUCT1,N)
> IDL> FOR I=0,N-1 DO STRUCTALL.PTR1=PTR_NEW(ALLDATA[*,I])
>
> Then I do what I need to do, and I'm done with it, so:
>
> IDL> UNDEFINE, STRUCTALL
>
> Except this doesn't free the memory, according to HELP,/HEAP.
>
> Since I'm not a real programmer, I still consider pointers to be kinda
> voodoo. Can someone patiently explain how I manually clean up the
> memory in this case?
>
> Thanks,
>
> --Edward H.
Hm, what's going on here. If I would do this on my PC the following
happens:
IDL> help,/heap
Heap Variables:
# Pointer: 0
# Object : 0
IDL> n=10
IDL> s={ptr1:ptr_new()}
IDL> help,/heap
Heap Variables:
# Pointer: 0
# Object : 0
IDL> sall=REPLICATE(S,N)
IDL> help,/heap
Heap Variables:
# Pointer: 0
# Object : 0
IDL> FOR I=0,N-1 DO SALL[i].PTR1=PTR_NEW(dist(i+1))
% Compiled module: DIST.
IDL> help,/heap
Heap Variables:
# Pointer: 10
# Object : 0
<PtrHeapVar1> FLOAT = Array[1]
<PtrHeapVar2> FLOAT = Array[2, 2]
<PtrHeapVar3> FLOAT = Array[3, 3]
<PtrHeapVar4> FLOAT = Array[4, 4]
<PtrHeapVar5> FLOAT = Array[5, 5]
<PtrHeapVar6> FLOAT = Array[6, 6]
<PtrHeapVar7> FLOAT = Array[7, 7]
<PtrHeapVar8> FLOAT = Array[8, 8]
<PtrHeapVar9> FLOAT = Array[9, 9]
<PtrHeapVar10> FLOAT = Array[10, 10]
IDL> undefine,sall
IDL> help,/heap
Heap Variables:
# Pointer: 0
# Object : 0
That's why I very like David's routines - they do always their jobs :)
Your problem might be that you don't correctly reference your pointer
array. You forgot:
STRUCTALL--->[I]<-----
So, maybe you allocate some memory each time in the loop, but then you
always overwrite the pointer(s).
Regards
CR
|
|
|
|
| Re: Memory Cleanup-- Messy structure [message #71463 is a reply to message #71382] |
Thu, 17 June 2010 23:47 |
rogass
Messages: 200
Registered: April 2008
|
Senior Member |
|
|
> I think it is proper for the times when IDL originated, when it was
> too expensive to do automatic garbage collection. IDL 8 has it (it can
> be switched off), though one could argue that it should not have taken
> so long.
When will IDL 8 be released? At the moment, Python comes closer and
closer...
Regards
CR
|
|
|
|
| Re: Memory Cleanup-- Messy structure [message #71467 is a reply to message #71382] |
Thu, 17 June 2010 17:10 |
Craig Markwardt
Messages: 1869
Registered: November 1996
|
Senior Member |
|
|
On Jun 17, 12:43 pm, pp <pp.pente...@gmail.com> wrote:
> On Jun 17, 12:45 pm, Craig Markwardt <craig.markwa...@gmail.com>
> wrote:
>
>> On Jun 16, 1:50 pm, Ed Hyer <ejh...@gmail.com> wrote:
>
>>> Since I'm not a real programmer, I still consider pointers to be kinda
>>> voodoo. Can someone patiently explain how I manually clean up the
>>> memory in this case?
>
>> I still think RSI/Kodak/ITT dropped the ball on the design of this.
>> IDL is a high level language, not a low level language - "4th
>> generation" right? IDL should be cleaning up messes like this, not
>> us.
>
> I think it is proper for the times when IDL originated, when it was
> too expensive to do automatic garbage collection. IDL 8 has it (it can
> be switched off), though one could argue that it should not have taken
> so long.
Uh, IDL "pointers" originated in IDL 5.0, ten years ago when all the
other scripting languages did automatic garbage collection.
I don't understand how it could be too expensive. The IDL interpreter
has precise knowledge and control of when variables go out of scope.
Dealing with cycles is its own problem, but basic garbage collection
should have been implemented from the start.
Now we are stuck with many years of maintaining meticulous pointer
allocation/deallocation discipline until IDL 5-6-7 are completely
obsolete.
Craig
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
