change dojox/mobile to reference module return values

[Douglas Hays]

converting calls to dojo.declare to use module return values

[Douglas Hays]

Kamiyama-san, please review the attached patch which changes the declare() calls to use module return values. The list of module dependencies for each file will be patched separately.

[ykami]

Doug, thank you for the patch. It looks ok to me.
Just to be sure, are we going to use declare instead of dojo.declare?

[bill]

(In [25874]) Convert dijit/_base to use module return values rather than dojo and dijit globals. Gets rid of references to dojo global, and some references to dijit global. Refs #13490 !strict.

[bill]

Oops, that comment above was for #13494.

Anyway, I looked over the patch too. Besides still using dojo.declare(), it's using the dojo global in lots of other places too, like for dojo.style(), dojo.addClass(), dojo.removeClass(), etc. I'll attach the script that I've been using (still in progress) to help do conversion of all these references.

[Douglas Hays]

(In [25876]) Refs #13490. Change declare() arguments in dojox/mobile to use module return values. !strict

[Douglas Hays]

added granular dojo dependencies and used module return values

[Douglas Hays]

Kamiyama-san, the 2nd patch needs to be reviewed and test very closely since it changes a lot. Also, please think about how using module return values will work with mobile's non-dojo code since you replace things like dojo.style, but the code is using domStyle.style. I'm not changing scrollable.js since it is redefining methods on dojo which is going to break baseless conversion.

[ykami]

Thank you for the 2nd patch. A couple of things I noticed so far.

• dojox/mobile/parser seems to be still accessing the global dojo.
• Some modules receive dojo/_base/kernel as "kernel", the others receive it as "dojo". Should they be consistent?
• _compat seems to be broken. Most of the test cases do not work on FF. Not investigated yet.

I will think about scrollable.js.

[bill]

FYI, my intention with renaming "dojo" to "kernel" was that "kernel" would only be used for the symbols defined in kernel.js, specifically experimental() and deprecated(). Then any remaining "dojo" references indicate an incomplete conversion.

[bill]

helper script to convert files to baseless (script not complete yet)

[Douglas Hays]

(In [25886]) Refs #13490. Add missing dojo/* dependencies and fully utilize module return values in dojox/mobile/. !strict

[ykami]

(In [25898]) Refs #13490 !strict Refactored scrollable.js to use module return values. Sorted the deps lists to be less error-prone. Fixed typos in the deps list of _compat.js. Removed unnecessary dependency on dojo.query.

[ykami]

(In [25912]) Refs #13490 !strict Added dependency on common.js to modules that really rely on common.js functions. Added existence check for this._cv to remove unnecessary dependency on common.js. common.js no longer extends _WidgetBase to add _cv.

[bill]

Hi, this conversion looks good but I'm still seeing a few global references:

• dojo.parser.parse (This one might be problematic because of how dojox.mobile.parser overrides the dojo.parser functionality, maybe you intentionally left it? Note though that the current code will break in 2.0 when there is no dojo.parser variable)

• dojo.xhrGet
• dijit.byId, dijit.byNode, etc. (it's referencing dijit as a global variable, should be requiring and then using dijit/_base/manager)
• dojox.mobile.app (again, perhaps this is intentional as that code is superceded by dojox.app?)
• ProgressIndicator.js is accessing dojox.mobile.ProgressIndicator, should access through local variable
• same for dojox.mobile._ScrollableMixin
• reference to dojox.mobile.SwapView should be just SwapView
• compat.js is referencing a bunch of dojox.mobile globals like dojox.mobile.RoundRect. I see how you are trying to, for example, check if dojox.mobile.RoundRect is loaded before patching it, but maybe it's better to load that file all the time. Since _compat.js doesn't even load on phones, it doesn't matter if it pulls in unneeded code. Also, with the AMD loader I don't think you are guaranteed that dojox.mobile.RoundRect loads before compat.js, so there's a race condition.

Also, common.js is loading but not using dijit._WidgetBase, so I think that dependency can be removed.

[ykami]

Thanks for finding them. I'll take a look. I may not be able to solve all of them, in particular, references to xhr and dojo.hash from dojox.mobile are a bit tricky ones..

[ykami]

In [26080]:

Refs #13490 !strict Removed global references as much as possible. Some of them, such as xhrGet and dojo.hash, cannot be removed yet though. Added all the necessary module dependencies to _compat following Bill's suggestion.

[ykami]

In [26089]:

Refs #13490 Updated the build profile to reference dijit.registry instead of dijit._base.manager.

[ykami]

In [26091]:

Refs #13490 !strict Removed most of the references to dojox.mobile. Hope I am doing right..

[bill]

In [26198]:

[25961] accidentally changed a margin-box to a content-box, although it doesn't seem to make any difference, refs #13490 !strict.

[bill]

In [30700]:

reduce uses of dijit variable, refs #13490 !strict