Re: [Modeling-users] Request for specific adaptors?

[Soaf <so...@la...>]

> I have recently felt that an adaptor to a non-relational db server would
> be very nice -- useful for simple applications (and installations of)
> and/or
> quick prototyping, but will allow that the same application be easily
> "scaled-up"
> to a relational server (with no changes to the data model).
> This does not equate to simple object persistence (e.g. any of several
> such schemes available, from pickle to zodb, etc.). It is also not about
> performance ;-!. It is about giving a very high-level separation between
> the application objects, and the back-end, that will also allow
> switching
> between SQL and non-SQL storage. I am in particular thinking of
> something like bsddb, with string values (rows) that are simply
> repr(dict).
> Then searching and filtering can still be achieved without modifying
> any client code (albeit very inefficiently!). The advantage is that
> small applications that do not do so much searching and filtering
> do not suffer very much, and they can be installed a lot simpler
> (requiring only to make sure to have write access to some directory,
> as opposed to setting up an sql server). But, the power of the model
> is still at hand, on which the application can build...
>
> mario


This is a good idea, since I usually don't use modeling for small stuff
as i need to use a big SQL server w/. PySQLITE is really a great stuff
i think my next dev about modeling will use this.

Perhaps we should look at gadfly ? (or ZODB + SQL )