From: Ivan B. <iv...@cv...> - 2011-11-05 16:30:04
|
On 11/04/2011 01:23 PM, Nathan Neulinger wrote: > I think this is a good approach. > > Do we have a "what's different in terms of support" document for trotl vs. otl and where we'd have to get to before the > old code would really be considered unneeded? Stuff like support for ancient oracle clients isn't a concern at all, but > if there is significant real functionality that is different, that would be good to know. > > -- Nathan > > The main advantage is support for wider set of the Oracle data types(like collections, long, anydata, xml, cursor, spatial will also be supported). It also uses more functions from OCI library, like OCIPing and OCIDescribeAny. These features are not used in Tora yet. Theoretically we can describe any database object and provide similar output as sqlplus have. OCIPing can be used to measure network round-trip time. In the future I would like add also support for AQ. For toConnection I would like to use something like PIMPL pattern as it is described in KDE policy guide http://techbase.kde.org/Policies/Library_Code_Policy#D-Pointers The real connection provider's implementation(subclass of toConnectionPriv) will we held in a shared library. This library will be loaded on demand. For example if OCI.dll is found then toOracle.dll will be loaded. When loaded this library will register itself in the toConnectionPriv factory. When using this approach we can distribute Tora on all platforms with build-in support for all the databases without confusing users. Ivan |