[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: object newbie



Maybe if I provided a few more details, it would be apparent whether I've 
missed the boat.

I have a widget that is used to read and graph data from a radar model
written in C.  I also use the widget to launch and control runs of this C
program on many machines across the network (thereby insulating me from
UNIX), as well as browsing and editing the input files and creating
additional runs according t a trade space.  This all was developed without
objects.

Based on discussion here a curiosity I decided to begin to "objectify" this
widget and began with the structure that contains "jobs", i.e. the path/name
of the executable, the path/name of the target file, the path/name of a dump
file containing what would normally go to the screen, a list of program
options, the lun being used, and IDs of some control widgets associated with
that job.  These widgets can be used to kill the remote job, look at the
dump and output files, and look at a status display containing most of the
information on the job object.  So to pass these fields into the display
widgets, I need access to most of the field values.

Is this a case where the object is inappropriate, or perhaps a case where I
simply haven't gone far enough, and more of the program should be
"objectified" in order to see the benefits.

 Chip

----------
In article <39991735.7D7D9C6F@dkrz.de>, Martin Schultz
<martin.schultz@dkrz.de> wrote:


> Pavel,
>
>    these are valuable comments. However, I disagree that objects are
> only useful for
> N >= 50 as you indicate. Especially when using inheritance, they have
> tremendous power
> even for N = 1 if you like some uniform access to "things" that belong
> to the same category
> but may all be a little different. Certainly you are right that
> structures still have their
> value, and you don't need an object for EVERYTHING in IDL. Yet, I have
> the feeling that
> your comment misses the point in answering Chip's question, and that
> this is probably
> because the question itself is flaky. I haven't pondered over this day
> and night, but
> I dare to say that you made something wrong when you need to know the
> fields that are
> stored in the object. Object oriented programming does require
> substantial twists of
> the sequential programmer's mind in terms of program structure. As
> always, the design of a
> program is the most important step in programming, but for objects it is
> even more
> important. Every object field (structure element) that you want to
> access from outside
> must be explicitely interfaced for a reason (encapsulation). All other
> fields should
> be regarded as private properties of the object for use only by the
> object itself.
> If you need access to object fields, the GetProperty, SetProperty
> convention also
> allows you to distinguish between Readonly and Writeonly access, simply
> by including
> or emitting respective keywords. Sure, OOP requires more typing than a
> classical program,
> but so far I still think I am rewarded in the end by a product that is
> much more powerful
> than a classical program of similar complexity. And if I had wanted to
> include all that
> power into a classical program, I probably would have had typed at least
> as much.
>
> Cheers,
> Martin
>
>
> Pavel Romashkin wrote:
>>
>> I have a feeling that Chip's question is not quite answered by the dump
>> truck full of technicalities the group readily poured out. It looks to
>> me that the question is, "my object is just a structure but it is hell
>> of a lot harder for me to use it than a regular structure".
>> I would say that if all you need is the data and there is one instance
>> of it, you do not need an object. The whole idea is that object is a
*reuseable*.
>> For example, you have one data processing task. You have a vector map
>> and an overlayed image. You want to be able to access the data for that
>> image only. Might as well write plain code for it.
>> Now, you have the same map, but 50 images that will be processed
>> similarly. In this case, if you write an object, then you can write the
>> "25 methods" only once, and very elegantly call those methods in your
>> widget event processing code. This way, you have intact data and a set
>> of actions (methods) you can perform on that data. And, if you happened
>> to aquire more bands of the image(s), you do not need to alter the code.
>> You just add more objects that use the same methods. On the other hand,
>> if you create a structure (or array of such), then you can't as easily
>> add another elements to it. Actions (code that modifies the data) tend
>> to become less organized (because they are not linked to data), and
>> event handlers become very complex.
>> This is not to mention inheritance, which simplifies dramatically
>> operations on data that are derivative or daughter to the main object class.
>> Cheers,
>> Pavel
>>
>> Chip Sample wrote:
>> >
>> > I must admit that based on comments from this list, I have experimented
with
>> > the object features of IDL for the past week or so, and have implemented
>> > them in a few places in a fairly big widget code I have written.
>> >
>> > My initial observations are that there seems to be less of a temptation to
>> > use common blocks when objects are used.  On the other hand my first
>> > impression was that an object is a structure whose fields can not be
>> > accessed until you write additional "methods" to get at each and every damn
>> > one of them.  So my object was littered with about 25 "methods" just so I
>> > could pry the data out of the object.
>> >
>> > I eventually came on a work around to write a "proto_object" with a method
>> > allowing you to pass a string containing a tag name which returns the
>> > contents of the field with that tag name.  This "proto_object" is inherited
>> > by all other objects I create just so I can use this method.  Along the way
>> > I found that the TAG_NAMES function in IDL doesn't work for objects so I
had
>> > to create one.  It basically copies the object structure into a regular
>> > structure so the TAG_NAMES can be used.
>> >
>> > Am I making this too hard?
>> >
>> > Chip
>
> --
> [[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[
> [[ Dr. Martin Schultz   Max-Planck-Institut fuer Meteorologie    [[
> [[                      Bundesstr. 55, 20146 Hamburg             [[
> [[                      phone: +49 40 41173-308                  [[
> [[                      fax:   +49 40 41173-298                  [[
> [[ martin.schultz@dkrz.de                                        [[
> [[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[[