Need advice on building an object container [message #66250] |
Mon, 27 April 2009 07:41  |
Paul Van Delst[1]
Messages: 1157 Registered: April 2002
|
Senior Member |
|
|
Hello,
I need a bit of advice on building a container for object.
I have an object definition, let's say 'RTS', that contains the simulated
top-of-atmosphere radiances for various satellites. I also have files that contains all
the instance data for a whole bunch of those objects. Those data are organised by the
sensor channel (i.e. different frequencies) and atmospheric profiles. So, in my
pre-object-code I would create a structure array after determining the dimensions,
something like
Rts = PTRARR( n_Channels, n_Profiles )
FOR m = 0L, n_Profiles-1L DO BEGIN
FOR l = 0L, n_Channels-1L DO BEGIN
Rts[l,m] = PTR_NEW({CRTM_RTSolution})
...read the current record...
ENDFOR
ENDFOR
But now I want to use objects (mostly as a learning exercise, but also to stop folks from
mucking about with the innards of the structure when they use it).
So, the structure definition of RTS doubles as the object defn also. What do people
recommend for reading the datafile? I see two options:
1) Use OBJARR(). Read dimensions and create the array and then fill it. However, this
would still require the user to create the array after inquiring the file for its
dimensions, and basically keep track of things.
2) Use a container. This is what I came up with this morning:
PRO RTSfile__Define
void = { RTSfile, $
Filename : '', $
FileId : 0L, $
INHERITS IDL_Container }
END
and, in the RTSfile::Read method I would simply read each RTS object from the file and do a
rtsfile->Add, rts
But this method doesn't preserve the basic [n_Channels, n_Profiles] structure of the data.
And, there's no indication in the RTSfile definition that this is a container for RTS
objects, rather than a generic container -- but I'm wondering if worrying about that is
just a distraction?
3) Sort-of combining (1) and (2) doing something like:
PRO RTSfile__Define
void = { RTSfile, $
Filename : '', $
FileId : 0L, $
n_Channels : 0L, $
n_Profiles : 0L, $
rts : PTR_NEW() }
END
where, eventually,
rts = PTR_NEW(OBJARR(n_Channels, n_Profiles))
and then fill in the object references to the RTS object.
So I was wondering how people would structure their code to handle this sort of thing. I
think it's a common enough paradigm that a pattern probably exists but I haven't found it.
Any info, hints, tips, appreciated.
cheers,
paulv
|
|
|
Re: Need advice on building an object container [message #66334 is a reply to message #66250] |
Fri, 01 May 2009 07:26  |
Paul Van Delst[1]
Messages: 1157 Registered: April 2002
|
Senior Member |
|
|
Mike wrote:
> Have you looked at the IDL_Container object? It could be inherited by
> your containers and used to handle all the bookkeeping.
Umm... yes. That's why I used it in my original example:
> PRO RTSfile__Define
> void = { RTSfile, $
> Filename : '', $
> FileId : 0L, $
> INHERITS IDL_Container }
> END
:o)
My problem is with the "packaging" of various levels of these objects. Or, alternatively,
translating my real world organisation of the data into something amenable to OOP
programming. My goal here is somewhat an academic exercise, but I want to be able to
easily change the definition of my base object without having to retool a bunch of code.
I think Ben's advice is what I needed. Just a pattern on which I can work with, or build on.
cheers,
paulv
|
|
|
|