> While it sounds neat, it's really not that much work (you only have to
> do it once) and it could be a lot of work (we'd have to include all
> kinds of special exemptions) and it would be for English only.
Well, it is much work, and as I mentioned in my post, because of bugs in Modeler,
I need to do it all the time the DB is changing, cause "overwriting" doesn't work well,
and only "first time" reverse engineering is delivering a nice and clean recognition/mapping.
> For example, what if you have an Address table? Does it make the Java
> class Addres? I suppose you could argue that the list of tables that
> would end in an "s" are fewer than those that wouldn't.
The idea is to use a NLP library that knows such things, and just use it.
IMHO this is not much work.
Besides, as I mentioned, it must not be perfect: it's good enough if it does for the 95% of
the words - it's still way more than doing the entire thing manually.
This archive was generated by hypermail 2.0.0 : Sun Nov 20 2005 - 10:15:06 EST