jeffreygg wrote:
Hey, Phil. Once you get your issue resolved (sorry I can't help with that) would you elaborate a bit for posterity on what the gotchas were in your conversion?
Hey Jeff,
Most of it really had little or nothing to do with LLBL per se, and most of it was easily correctable. So these are more general gotchas then LLBL-generated DDL gotchas.
I should also say that I may not remember the whole trial and error process accurately, as I had to get the new version to compile (not run) as fast as I could, and once I got something working I never really looked back.
-
The issue with the quotes, as described in my original post.
-
Some names were truncated down to 30 characters, and some were not. Tables names were not, if I recall correctly. Foreign Key constraint names were truncated, but oddly not always properly. One example:
"FK_lead_activity_document973AD3D7"
is 33 characters. Not sure if it ignored the "FK_" when generating the name? I should have documented this as a possible bug, but (again) I was really pressed for time.
-
Stored procedures in Oracle can only return DataSets, not DataTables, regardless of how many resultsets they return, so interfaces for SPs had to be refactored.
-
I was concerned with having to change a bunch of interfaces for SPs again for each retreival procudure, since I would have to use a ref cursor as an IN/OUT parameter for each one. But as it turns out, LLBL handles this for you.
-
Of course, Oracle sequences had to be created for all SQL IDENTITY columns, and then had to be mapped in the LLBL designer.
-
I had to use the 10.1x ODP driver instead of the 10.2x driver. I was never able to figure out why I couldn't get the newer driver to work.
That's all I can think of at the moment. If I think of anything else I'll post it.
Overall, it was a pretty easy conversion. I actually converted from Version 1 of LLBL to version 2 at the same time, and got everything to compile in about 4 hours, including writing a script in SQL to generate Oracle SP stubs.
Phil