> I have to say that randomly creating DB schemas and then reengineering
> them is probably the weirdest, but also coolest thing I've ever heard of.
> Welcome Andriy! :-)
> Keeping this regression suite in parallel to the regular unit tests is a
> very good idea. A simple 'regression' ant target would be nice to have,
> later.
>
You know, it turns out such an application is a decent bug finder and helps
with metering performance related stuff too. You can find the code in
src/regression. It does not work with official Cayenne but it will next week
after the new integration version is checked in. Right now I am adding tests
of reflexive relationships to the regression app.
> I might be missing something but one thing I'd like to see is a kind of
> pluggable PKFactory that allows me to quickly provide new per-entity PKs
> without exposing the attribute on the DataObject. Something along the
> lines of EOF's EOAdaptorChannel.primaryKey(s)ForNewRowWithEntity. Is this
> what you have in mind, too? Especially for DBs that can handle the special
> unique-24-byte-PK the performance is just so much better and solves many
> PK related race conditions with multiple writers.
>
Principally, yes, the goal was to introduce configurable optional key
generators (or factories) per DbEntity so that one could define name, cache
size, and type in a data map and everything else would be handled
automatically based on the type of the specified generator. The format is
documented in the data-map.dtd and, hopefully, is suitable for definition of
various kinds of key generation. By now Oracle sequences have been
implemented due to lack of time. But as time allows I will continue on that.
Andriy.
This archive was generated by hypermail 2.0.0 : Sun Feb 02 2003 - 12:39:01 EST