[Nitro] Roadmap 0.50

[Jonathan Buch]

Hi,

> Not fed up.  Just absorbed on other projects.  I'm still around, and Og
> especially is high on my list of projects.

thats good!


> I was destroying the db after each test in order to make sure that each  
> spec is independent.  The classes are one thing, entries in the db are  
> another.

> Yes, this'll successfully restrict the class defs to the context they're
> part of.
>
>> and could the quick_setup  be revised to reuse a single Og connection?
>
> Maybe, but: the real problem, IMO, is that storing stuff to a database is
> necessarily a stateful operation.  So, if the DB isn't destroyed between
> specs, the specification isn't really valid, you know?

Which is why many old test cases only use one big test_ method.  I agree
that this goes against the principle of making tests as small as possible
and only touch one 'concern'.
Allright, I give up for now, lets only make specs and not complicate them
further.  :P

> One possibility that I like would be to use SQLite in memory mode, since
> that'd be blazing to setup and teardown - except that SQLite doesn't  
> seem to like to drop and database unless all the connections on it are
> properly closed.

One more thing to take care of, but also nothing which would prevent the
next release...  Reminds me to go look at the postgres drop db, I think
there were some problems..

Jo

-- 
Using Opera's revolutionary e-mail client: http://www.opera.com/mail/