Provide a generic interface and base class for getting, setting, and watching for property changes (with getters and setters) in a consistent manner

[Kris Zyp]

This ticket proposes the addition of a dojo.Stateful class, providing the ability to monitor stateful objects in dojo with a watch()}} method. The {{{watch() method would be based on Mozilla's Object.prototype.watch (1):

obj.watch(propertyName, callback); // monitor a property

or

obj.watch(callback); // monitor any property change for that object

The two primary types of objects that I would like to be able to monitor are widgets and data store objects. We have recently discussed using a different data/object store approach whereby data store objects could be provide a consistent set of methods, and watch() could certainly be one of them. dijit._Widget will extend from dojo.Stateful and thus deliver notifications to listeners for any attribute changes. The ability to watch an object would be discerned with feature detection (the presence of obj.watch), but would hopefully be available at least on all widgets (although user objects could implement this ability as well).

One of the primary motivations for this design, is would enable building functionally reactive "live" templating engines or data-bindings on top of Dojo. The concepts of functional reactive design (or data-bindings) have been demonstrated in other projects like Flex, Flapjax (2), and Fin (3). There recently has been interest expressed in developing a template language with these type of data-binding capabilities that could run on Dojo, where one could write something like:

<div>${obj.foo}</div>

or

<div>${obj.get("foo")}</div>

And the variable/object property would be evaluated when the template is executed, and the templating engine would then automatically "watch" the obj's foo property for any changes and automatically rerender the template when the value changes. This is a powerful concept because it can remove a significant amount of user code for event propagation and rerendering.

I think another beneficial pattern for clarity of code with code-based attribute access is the use of get()/set() pattern for code-controlled property access. We have discussed this before, but I think this is much clearer the current widget attr() that handles both gets and sets. watch()/get()/set() would make a clear easy-to-use pattern for interacting with widgets.

(1) ​https://developer.mozilla.org/en/Core_JavaScript_1.5_Reference/Objects/Object/watch

(2) ​http://www.flapjax-lang.org/

(3) ​http://ajaxian.com/archives/fin-self-updating-template-language

[Kris Zyp]

dojo.Stateful

[Kris Zyp]

dijit._Widget

[Eugene Lazutkin]

Reformatting.

[Kris Zyp]

(In [21666]) Addition of dojo.Stateful for get/set/watch objects, refs #10839

[Kris Zyp]

(In [21667]) Addition of dojo.Stateful for get/set/watch objects, refs #10839 !strict

[bill]

(In [21676]) just fixing doc typo, the first parameter to watch() is optional, not the second, refs #10839.

[bill]

(In [21677]) just fixing doc typo, the first parameter to watch() is optional, not the second, refs #10839.

[Kris Zyp]

(In [21687]) Reverts _Widget extending dojo.Stateful, and fixes connection points to attr method, refs #10839 !strict

[Kris Zyp]

(In [21689]) Changes grid handling of get() to be compatible with _Widget, refs #10839 !strict

[Kris Zyp]

(In [21754]) Support hash based sets, refs #10839 !strict

[Kris Zyp]

(In [21755]) Added improved get/set docs, updated tests, and support hash based sets, refs #10839 !strict

[bill]

(In [21764]) fix syntax errors from [21755], refs #10839

[Kris Zyp]

(In [21779]) Fix Tree + JRS test, refs #10839

[bill]

(In [21781]) Rollback strange change from [21755], was this checked in by accident? Buttons do not support width/height setting. Refs #10839.

[bill]

(In [21782]) API doc for hash based sets, refs #10839

[bill]

(In [21783]) Update API doc (mainly for deprecated methods), redirecting to use set()/get(), rather than redirecting to use attr(). Also added API doc for set() of hash based sets. Refs #10839 !strict.

[bill]

(In [21784]) Conversion of Widget internal this.attr() calls to this.get()/this.set() calls, refs #10839 !strict. (There are lots more calls left though.)

[bill]

(In [21786]) Need to connect to set() not to attr(), in case set() is called directly. Refs #10839 !strict.

[bill]

(In [21787]) Conversion of remaining dijit attr() calls to get()/set() calls, refs #10839 !strict.

[Jared Jurkiewicz]

(In [21820]) Missed a conversion of an attr call. refs #10839

[Jared Jurkiewicz]

(In [21821]) Converting a bunch of dojox Editor plugin calls to yse get/set instead of attr. \!strict refs #10839

[Jared Jurkiewicz]

(In [21839]) Fixing usage of attr to get/set to remove deprecation warnings. refs #10839

[Kris Zyp]

(In [21858]) Reverted change to test tree data structure, Fix Tree + JRS test, refs #10839

[bill]

(In [21864]) Only print attr() deprecation warning once per calling function, refs #10839 !strict.

[bill]

(In [21957]) attr() --> get()/set(), refs #10839 !strict

[bill]

(In [22068]) attr() --> get()/set(), refs #10839 !strict

[bill]

See also [22101]: Figure.get -> Figure.getAnnotator to avoid conflicting with the new set/get style

[bill]

See also #11105.

[Kris Zyp]

(In [22138]) Fixes |this| for callbacks, refs #10839

[bill]

(In [22237]) Fix problem in attr() deprecation warning when arguments.callee.caller is null. Fixes #11203, refs #10839 !strict.

[bill]

I filed #11251 for the _Widget.watch() feature. I think you can close this ticket now?

[bill]

(In [22543]) fix comment, refs #10839 !strict

[bill]

(In [22773]) Refactor TitlePane so set("open", ...) is main interface to open/close, rather than toggle(). This also moves the CSS code (including _setCss() call) from postCreate() into _setOpenAttr() thus doing no CSS changes in postCreate(). Refs #10839, #11635 !strict.

[bill]

(In [22834]) update comments from attr() --> get()/set(), refs #10839 !strict

[Kris Zyp]

(In [23669]) Make a copy of callbacks before iterating to avoid iteration conflicts, refs #10839

[bill]

In [30263]:

fix order of arguments to doh.is() (should be: expected, actual), refs #10839

[Bill Keese <bill@…>]

In 59323f72ec1b98270fec565da6d30efd96fd7fbe/dijit:

Error: Processor CommitTicketReference failed

Unsupported version control system "git": Can't find an appropriate component, maybe the corresponding plugin was not enabled? 

[Bill Keese <bill@…>]

In d71e794118448491b18bf92a7643f4cda544162c/dijit:

Error: Processor CommitTicketReference failed

Unsupported version control system "git": Can't find an appropriate component, maybe the corresponding plugin was not enabled?