dijit/registry, dijit/a11y, dijit/WidgetSet

[bill]

Split off dijit registry, a11y related functions into separate AMD-friendly files:

• dijit/registry
• dijit/WidgetSet (deprecated class)
• dijit/a11y (tabindex related functions)

dijit/_base/manager.js remains for backwards compatibility.

[bill]

(In [26050] Splits out tabindex related functions from dijit/_base/manager.js to separate dijit/a11y.js).

(In [26052]) Refactor dijit/_base/manager.js, putting common registry related methods into dijit/registry.js, and WidgetSet stuff in WidgetSet.js. WidgetSet is essentially deprecated, and dijit/_base/manager in general is unneeded; most widgets can just require dijit/registry to get the functions they need (byId(), byNode(), etc.) WidgetSet.js is still automatically loaded in legacy mode, when async:false, just in case some old app was depending on features like dijit.registry.forEach().

(In [26061]) dojo.registry.filter(...) ->dojo.forEach(dojo.registry.toArray(),...)

(In [26071]) Make dijit.byId() etc. available in legacy (non AMD) mode even when dojo/_base/manager not explicitly loaded.

[bill]

In [26079]:

update dojox.calc to use dijit/registry, and _TextBoxMixin, refs #13603 !strict

[bill]

In [26082]:

byte trimming, refs #13603 !strict

[ben hockey]

Replying to bill:

(In [26071]) Make dijit.byId() etc. available in legacy (non AMD) mode even when dojo/_base/manager not explicitly loaded.

i'm having problems with this commit and since i don't fully understand all that its doing (or why its doing it the way it does) i'm not sure how to try and fix it.

the problem: the globals (dojo, dijit, dojox) can be remapped and the "naked" reference to the dojo global fails.

so, a better test might be if (typeof dojo !== "undefined" && dojo.ready && !dojo.isAsync) { ... }.

however, i'm not sure why you aren't just pulling in dojo/_base/kernel and using it to check for !dojo.isAsync in this case. that would also handle the case when the globals are remapped whereas the typeof check wouldn't work as intended if the globals are remapped. of course, you'd also need to add dojo/ready as a dependency to use it.

what am i missing that means this wouldn't work?

[bill]

I could have pulled in dojo/_base/kernel, but I was just trying to minimize the amount of code, and I guess also I was trying to remove dojo/_base/kernel from the dependency lists of dijit modules, since dojo/_base/kernel in the dependency list usually indicates that a file wasn't fully converted to granular dependencies.

I didn't consider the remapping issue though. And I didn't consider that the condition:

if(dojo && ....)

will fail when dojo isn't defined because has("extend-dojo") is false.

So, feel free to update those modules (dijit, fx.js, dojo/dnd/Source.js, etc) to pull in and use dojo/_base/kernel (note that the dijit dependencies are in alphabetical order, and it always maps dojo/_base/kernel to the variable "kernel"), or alternately to change the if() condition as you listed above.

[bill]

Rawld took the second approach for _WidgetBase.js in [26745], but the other files are still broken.

[ben hockey]

i'm going to track this in #14033 - marked with high priority so it makes the 1.7 hot list.

[ben hockey]

btw - my preference will be pulling in dojo/_base/kernel rather than what was done in r26745.

[bill]

In [26781]:

Include WidgetSet from dijit/_base.js layer module, since it used to be part of base. Refs #13603, #14043 !strict.

[bill]

In [26798]:

add deprecation comment, refs #13603

[bill]

In [27065]:

Add tests for new dijit/registry module. Will keep dijit/tests/_base/manager.html around until 2.0, for testing the old API. Refs #13603.

[bill]

In [27068]:

Move a11y test to top level since dijit/a11y is now top level, refs #13603.

[bill]

In [27069]:

Missing dojo/domReady! dependency led to intermittent test failures, refs #13603.

[bill]

In [27310]:

[27068] commented out some code by mistake, putting in back, refs #13603.