Re: OOP and load/save

[Robert T. Short]

John W. Eaton wrote:
> On 24-Apr-2009, Robert T. Short wrote:
>
> | BTW.  IMHO, I am not at all sure the notion of converting the class to a 
> | struct when the structure is unknown is valuable.  I realize that it is 
> | non-compatible, but this is one of those cases where it simply may not 
> | be worth the effort.  Returning a meaningful error would be just fine, 
> | at least for my purposes.
>
> Suppose you have a data file that contains some objects and some other
> data, and the constructors for the objects are unavaiblle.  If you get
> an error when reading the objects because you don't know how to
> construct them properly, then that renders the entire data file
> unreadable.  So continuing in some way with reading makes sense.  I
> guess it would also be OK to skip the objects instead of converting
> them, but if they are converted to structs you can at least examine
> the data they contain even if they aren't useful as objects.
>
> jwe
>
>
>
>   
Good point.  Basically, it seems to me that the MATLAB way of doing it 
is flawed, so some way of recovering is better than nothing at all.

Even if you could reconstruct the class (e.g., the parent list was part 
of the saved data), if the user had changed the class structure since 
saving it, the results would be a downer anyway.  So, you are probably 
right.  Returning the data so the user can see what he(she) did at least 
allows for sensible debugging and possibly a way of correcting the 
error.  In fact, the more I think about the more I think you have the 
right of it.  If I save some valuable data, then make changes, and then 
read the old data, I can probably make a converter that will make the 
old class match the new class.  This could be very important.

So, tomorrow morning I will spend a couple of hours looking at this.  Is 
there something you would particularly like me to look at?

Bob