Le 18 janv. 2011 à 14:35, Rob McBroom a écrit : > On Jan 18, 2011, at 1:03 AM, Patrick wrote: > >> A rewrite is definitely the best plan, but that's not to say people haven't >> thought about it already. > > Alcor started [a rewrite][1] and there's [another][2] underway. (Not sure if > that second one is supposed to be talked about, but I figure I'm allowed if > it's on GitHub.)
Oh, you're allowed, I was showing my current plans to Ankur Oberoi after a little "development" meeting we had ;-). But this is only preliminary stuff, and it's using MacRuby which I just discovered and which really speeds up Cocoa development. But my mind isn't made up on the MacRuby thing. > Just pointing those out for the benefit of anyone who's thinking about > starting from scratch. > > I know the people having this discussion are pretty familiar with > Quicksilver, but still I have to wonder if you really understand the glut of > features Alcor was able to implement. It almost seems impossible that one > person did all this. Are you sure a rewrite is realistic? I'm pretty sure a rewrite is realistic (:-P), it's just that personally I'm time-crunched on other things atm. Right now I have (another) ObjC only rewrite with a little part of the basic model stuff "working" (as in, you can create objects (files, urls, text, collections) from a *program*) and tested. It just lacks some more model classes (actions, commands), and... also everything else :-S. I just want to point out that I would prefer the main development repository to be https://github.com/quicksilver/quicksilver. It's just that my notification queue is getting messy, and the one there actually has Alcor's rewrite removed (since IMHO they don't belong to the same repository, having their history separated). Now, I do have some plans on continuing my rewrite, it's just that I'm at a loss at how to handle things like the Catalog in a CoreData world w.r.t performance/ease-of-use in such a flexible object world. And the catalog is the thing that's used for providing search results to the interface. I have a test project ongoing to check that performance with CoreData is acceptable enough to warrant its use, but I'm not done yet. The plugin system (https://github.com/tiennou/qs-elements, extracted from Alcor's rewrite because I guess it's general-purpose enough to warrant such a move) is mostly working, though it lacks updating and some kind of repository-like function like current QS has, which are also on my rewrite queue but lower that the Catalog thing, since that's the thing preventing my rewrite from advancing. IMHO, the main problem with the current code is that it predates stuff like KVO/KVC, Core Animation, Core Data, which means a lot of code (even whole classes on the Core Animation POV) could be removed, which adds up to more files and more bugs :-(. If you're interested in helping development, I'm planning to publish a basic rewrite with some kind of Catalog thing and documentation (BTW, thanks Jordan Kay for the link to AppleDoc, I guess I'll go with that) as soon as I feel it's a good start, so maybe during February. Cheers to all, and thanks again for the support ;-). -- Etienne Samson [email protected]
