valeriy said: So, what prevents you from using TROComplexType?
In fact - for now nothing.
But because i’m still searching for best multiplatform solution i have only one doubt.
For now. Let me explain.
We are searching for solution/framework that allow change data between server
implemented on .NET, Win32 and java platform and clients implemented on .NET, Win32, Android and iOS.
RO SDK and dataabstract covers most of our needs.
Dataabstract mostly on dataset oriented apps (like delphi clients), RO on object oriented apps.
When we use RO methods for changing data between client and server we can use simple data types (int, string, etc). But in real live objects (clasess or structures) are
more usefull.
Classes are universal, we can use inheritance, annotations (please look at my question
http://connect.remobjects.com/discussion/1738/attributesannotations-in-ro-sdk-structures), we can serialize it in any possible format, etc.
In demo apps, we can design classes by hand.
But in complex solutions we should use some design tools like (Visual paradigm, etc).
These tools (most of them) can generate class definition code for many platform
(#C, delphi, java, js …) and some of them can generate code for hibernate, nhibernate,
entityframework, etc.
But all that tools generate code on platform specific base class.
Not on some special like TComplexType.
This is first point of my doubt.
Second point:
Before Delphi 2010 RTTI was very poor. But in Delphi 2010+
RTTI is much more powerfull. If you look on google code you will find a lot of
projects creating frameworks for ORMs, serialization, etc. All based on pure PODO classes.
So in very near future using PODO in Delphi is going to be like using
POCO or PODO in .NET & java.
So creating all Data structures derived from TComplexType is not tacticaly good.
I believe RO SDK has many pros against datasnap. For now.
I’m afraid that not for long !
But this is topic for another discussion.
Any different opinion will be preferred !
Thanks for advance.
Edvin