port connect() + tests into core

[alex]

as per the roadmap document and refactor.txt, port dojo.event() and tests into Core. Move dojo.event.* to dojox.advices.

[sjmiles]

(In [7988]) First draft of new connect/event stuff. Refs #2641

[alex]

(In [7994]) API location refacotoring, style updates, and reformatting. Refs #2641

[alex]

(In [8235]) committing Scott's new keyboard handling work + tests. Refs #2641. Still needs more unit tests.

[alex]

(In [8249]) I'm a freaking idiot. My apologies to scott and adam for screwing things up for both of them. Refs #2641

[sjmiles]

(In [8264]) Event.js: all browsers: set keyCode to 0 when charCode has a value, norm CTRL-BREAK to CTRL-c where possible, IE: use native keypress for ESC and ENTER refs #2641

[guest]

patch for issue introduced in rev.8264: evt.keyCode is read-only in IE if key is ctrl or shift, thus try/catch is required, as already implemented elsewhere in the same file

[sjmiles]

(In [8306]) Remove vestigal argument to "dispatcher". Refs #2641.

[sjmiles]

(In [8308]) Make sure stealth onkeydown listeners are disconnected with their progenitors. Fixes last known missing requirement for event.js. Refs #2641

[sjmiles]

(In [8402]) Another round of revisions to connect and event. Refs #2641. *connect*: had to separate argument normalization from the main functions (hopefully this will aid the doc parser since there is only one "connect/disconnect" pair now, and overrides are done to private "_connect/_disconnect" functions. *event*: eliminated confusing 'clean' methods and moved a key function out of the closure for leak prevention. *disconnect*: changed disconnect so that it only takes one argument: the handle returned from connect, simplifying client code that needs to manage connections. Updated code in various modules to reflect this change.

[sjmiles]

(In [8403]) Updated code in various modules to reflect changes in Dojo r8402. Refs #2641. *disconnect*: changed disconnect so that it only takes one argument: the handle returned from connect, simplifying client code that needs to manage connections.

[sjmiles]

(In [8404]) Updated code in various modules to reflect changes in Dojo r8402. Refs #2641. *disconnect*: changed disconnect so that it only takes one argument: the handle returned from connect, simplifying client code that needs to manage connections.

[sjmiles]

(In [8424]) Fine tuning in connect. Redo IE keyboard magic in event.js so it actually works. Update docs. Refs #2641.

[sjmiles]

Diagram I dashed off to explain the onkeypress/onkeydown trick on IE. For posterity.

[alex]

(In [8431]) style guide conformance for changeset 8402. Refs #2641

[sjmiles]

(In [8457]) Fix bug (x2) setting stealthKeyDown flag. Thanks Doug. Refs #2641.

[guest]

Concerning changeset 8457 done by sjmiles:

I do not understand why "!kd.listeners" is present in the condition on line 231. May be it should be simply

if(!kd||!kd._stealthKeydown)

Best regards,
Konstantin Kolinko

[sjmiles]

Someone could have replaced the Dojo dispatcher at onkeydown (kd) with some other function. In this case, _stealthKeyDown is true, but kd.listeners is false.

Another scenario is that someone clobbers the Dojo dispatcher, then connects to onkeydown, then connects to onkeypress. We don't actually trap this case. I suppose we could write a more robust test to see if a stealth listener is actually there, but it seems like an awfully unlikely circumstance.

Thanks for studying this code.

[sjmiles]

(In [8474]) Fix test that was using the console instead of the test object. Refs #2641.

[guest]

It looks like the new connect() is changing "onevent" to just "event" in Mozilla. This works for DOM events, but it doesn't work for connecting normal object methods.

say I have a object { onrender: function(){} } dojo.connect(obj, "onrender",...) will try to connect to "render" in Mozilla and "onrender" in IE...

[sjmiles]

It should only munge the names if it detects that 'obj' is a node, in which case it assumes you are trying to connect to a DOM event.

If you have a test case where it is doing this and 'obj' is not a node, please post the particulars.

If 'obj' is in fact a node (or node-ish) but you want to connect to a custom event (not a DOM event) pass true for the fifth parameter (dontFix) to connect.

[guest]

Thanks! I just realized that my object has a method named addEventListener...

The node-ish check (nodeType, attachEvent, addEventListener) probably should be documented near connect(), it wasn't obvious at first sight.

[alex]

I'm gonna mark this closed. We can file new issues under separate cover.