2008-10-13:
[9:32] <hannesw> So what are the preferences out there regarding regular expression based url resolving vs. an object graph/name convention based scheme?[9:33] <hannesw> I just checked out cherrypy last week, and it's probably the closest to Helma 1 in terms of url resolving you can get.[9:34] <hannesw> I do think the _action based convention had some merits, or something like setting a "action" or "exposed" flag on functions...[9:35] <hannesw> But something I don't want anymore is resolving url paths by object references. That just mixes up things that shouldn't be mixed.[9:35] <hannesw> Any comments?[9:37] <zumbrunn> I like the object graph/name approach for much of what I do[9:37] <zumbrunn> but over all, I agree the regular expression based url resolving offers a better foundation[9:38] <zumbrunn> the object graph based behavior can always be added on top of that[9:39] <zumbrunn> the two approaches aren't mutually exclusive, in by opinion[9:40] <hannesw> no, they're not.[9:40] <zumbrunn> in other words, +1 for using regular expression based url resolving as the default way of going about that[9:41] <hannesw> IMO the object graph approach is viable if we also add some explicit expose mechanism.[9:41] <hannesw> similar to explicit __export__, maybe:[9:41] <hannesw> __expose__ = ["index", "list", "someSubModule"][9:41] <hannesw> ?[9:44] <zumbrunn> yes, something like that would be nice[9:44] <zumbrunn> but even if it needs to be custom coded whenever an object graph based approach is needed, that would be fine[9:47] <zumbrunn> btw, regarding the dropping of Object.dontEnum.... it probably will not make sense for __defineProperty__ to actually make its way into Rhino either[9:48] <zumbrunn> since there are now other proposals for ES3.1 that have been discussed in a lot of details[9:49] <zumbrunn> Object.defineProperty and Object.defineProperties[9:49] <zumbrunn> and some other methods, I believe[9:50] <zumbrunn> (they do not work exactly like __defineProperty__ does now)[9:50] * zumbrunn looking for where that was discussed...[9:52] <zumbrunn> http://wiki.ecmascript.org/lib/exe/fetch.php?id=es3.1%3Aes3.1_proposal_working_draft&cache=cache&media=es3.1:rationale_for_es3_1_static_object_methodsaug26.pdf[9:54] <zumbrunn> If I remember correctly, the way it is specified there is now pretty much agreed between the various ecmascript stack holders[9:58] <zumbrunn> apparently, the 9/22 draft even contains "specs" for it:[9:58] <zumbrunn> http://wiki.ecmascript.org/lib/exe/fetch.php?id=es3.1%3Aes3.1_proposal_working_draft&cache=cache&media=es3.1:tc39-es31-draft22sep08.pdf[10:00] <zumbrunn> (there is nothing wrong with having __defineProperty__ as a dontEnum replacement in helma for now... I'm just saying, it probably doesn't make sense to push for it going into Rhino)[10:02] <hannesw> thanks for that info, chris[10:03] <zumbrunn> regarding __expose__.... how would the "someSubModule" part work? not sure what you meant with that[10:03] <zumbrunn> "someSubModule" would be the name of a module that would handle the request somehow?[10:03] <hannesw> that would be the name of a imported module that is also exposed[10:04] <zumbrunn> what inside the module would be exposed?[10:04] <hannesw> yep, for a request /someSubModule/index the lookup would then proceed to look for "index" in someSubModule[10:04] <hannesw> basic delegation[10:04] <zumbrunn> I'm confused because it doesn't resolve to a function[10:04] <zumbrunn> ok, I see[10:06] <zumbrunn> that's kind of cool, actually :-)[10:06] <hannesw> yep[10:08] <hannesw> ok, the defineProperty() proposal is actually very similar to what we have, except the flags are packed into a {writable: ..., enumerable: ...} descriptor[10:09] <hannesw> off for lunch...
In the channel now:
Logs by date: