Re: Questions for eventual Python module

On Mon, Aug 26, 2002 at 05:52:16AM +0300, Timo Sirainen wrote:
> On Sun, Aug 25, 2002 at 04:31:24PM -0700, Manoj Kasichainula wrote:
> > I don't have irssi-signal handling working quite yet. The main thing in
> > the way is that I need a good way to recognize the type of each
> > structure pointer (server, window, reconnect, etc.) and map it to the
> > python wrapper object I've written for it. I haven't quite parsed out
> > yet how the Perl module does this yet; what's the easiest way to go
> > here?
> 
> Perl uses docs/signals.txt and generates perl-signals-list.h from it, using
> get-signals.pl. You'd probably do something similiar to that.

Hmmm. So modules that add their own signals couldn't have their signals
processed by scripts?

Maybe in the future, each _REC could include a field with a magic number
(or a string pointer) indicating the type. Then signal handling code
could look at the magic numbers and convert them appropriately. I only
suggest this because I know you're pondering a rewrite :)

> > What kind of relationships (inheritance, delegation, etc.) should I try
> > to set up between the various irssi objects? For example, I haven't
> > quite figured out whether SERVERs should be considered subclasses of
> > SERVER_CONNECTs, vice versa, some form of containment/delegation, or
> > none of the above. Any advice here would be appreciated.
> 
> They're not inherited too much. Mostly just IRC (and other protocol) objects
> inherit the core objects. Then CHANNEL_REC and QUERY_REC are inherited from
> WI_ITEM_REC. I'm not sure if there was others. You should be able to see
> these by looking if there's #include inside structure definition.
> 
> SERVER_REC contains connrec which points to SERVER_CONNECT_REC, but I don't
> think it should be considered inherited or anything.. or .. um.. well.. :)

That was the impression I had :)

> Looks like my perl code just merges the server and server_connect into one
> perl object. I probably did it just because it was easier that way to use,
> guess you should too :)

Hmm, ok. This might actually make a lot of sense; I'll think about it.
Thanks.