YUI 3 :: {79} Tickets Addressed in YUI 3.4.0 PR3

Anim - ability to work with non-node objects

Currently, anim only works with node instances or elements that can be converted to node instances. Enabling it to work with any object that has attributes would be useful. (e.g. shape instances in the Graphic api)

Chrome fires multiple popstate events + race condition with two concurrent .save()'s

It has to do with pushState() being called in the head.

from IRC:
rgrove Chrome's popstate behavior does differ from other browsers, but this is a new wrinkle I hadn't seen before. Hmm.
rgrove Actually, the bug occurs with the code in the body too, but not if you wait until after window.onload.
rgrove Basically if you call pushState before the first popstate event for the pageview (and Chrome always fires one, whereas other browsers only fire for pushState pageviews), then the bug occurs.

part 2:
Race condition.
rgrove The popstate event doesn't actually include any information on the new URL, so we have to manually inspect window.location to figure out what the URL has changed to. Unfortunately, window.location isn't guaranteed to be updated before popstate fires, so we have to use an async timeout to inspect the location *after* the popstate event (and everything else in the event queue) are finished...

Ref url http://jsfiddle.net/triptych/zpLNe/

model-list doesnt sort properly on refresh when using comparator

when using a custom comparator function, the models are not sorted correctly after refresh.

destroying a model does not remove it from all lists

when a model belongs to multiple lists, destroying it does not remove it from all its lists

the destroy method in model uses array.each to iterate over its lists property, when list._detachList is called, it splices the models list array and the array.each loop has now skipped an item.

need to make a copy of the items array before looping in model.destroy.

ModelList subclasses can't host their own attributes

ModelList's 'get' method is provided by ArrayList.addMethod, which aggregates calls to get for all items, any own ATTRS are inaccessible.

Something like this should do:



Y.ArrayList.addMethod(ModelList.prototype, [

    'get', 'getAsHTML', 'getAsURL'

]);

ModelList.prototype.get = (function (getFromChildren) {

    return function (attr) {

        return this.constructor.ATTRS[attr] ?

            Y.Attribute.prototype.get.apply(this, arguments) :

            getFromChildren.apply(this, arguments);

    };

})(ModelList.prototype.get);

But it's late, so maybe that is a terrible idea.

Adding the same model twice to a model-list always gets you an error.

I was trying to have a list that silently ignore any subsequent re-insertion of a model and did this:
[code]initializer: function() {
this.on('add', function myDupCheck(e) {
if (this.indexOf(e.model.get('id')) > -1) {
e.preventDefault();
}
}, this);
}[/code]
But that myDupCheck function never get executed, because the code that checks for duplicate models throws an error before firing the "add" event.
I tested this behaviour in 3.4.0pr1, but I checked github and saw the same code.

Split README into HISTORY.md and README.md

This ticket is assigned to AsyncQueue because it is first in alpha order of my components, but this ticket applies to all components.

Use expanded request string as cache key instead of just query in AutoComplete

Discussion from IRC:

"(04:23:43 PM) brian_endpoint: rgrove: me again :-)... it doesn't appear there is a way to not cache queries, and it seems like the request after the request template has been used should be the cache key rather than query itself
(04:24:07 PM) brian_endpoint: for instance I'm using requestTemplate to append a value from another field which may have changed even though my input node did not
(04:24:07 PM) rgrove: Talking about browser caching here?
(04:24:18 PM) brian_endpoint: no, _createIOSource caching
(04:24:29 PM) rgrove: Oh. Let me pull up the source and refresh my memory.
(04:25:14 PM) brian_endpoint: https://github.com/yui/yui3/blob/master/src/autocomplete/js/autocomplete-sources.js#L80
(04:26:01 PM) rgrove: Got it. Hmm.
(04:26:17 PM) rgrove: Yeah, I need to change that to use the full request as the cache key.
(04:26:19 PM) brian_endpoint: my use case is that I have a countries pull down and zip code entry field, I'm using requestTemplate to put the country value from the pulldown on the query string with the zip code entry, if you change the country I fire request, but if zip code hasn't changed (the query) I get the cache"

[Pull Request] - Fix bug #2529662 Modify flush method to optionally take a request object Modify flush defaultFn to check for event args & then either flush everything or single item Abstract out finding an element in the cache _entries array to _position method Modify retrieve to use _position for attempting to find cache entry

Update to Y.Cache to let .flush() take an item to flush from the cache.

Multi-pane calendar template support

sprint tracking ticket

Unit tests for single calendar rendering

Rules matching unit tests

Rules matching implementation

Custom header renderer

double pane navigable, selectable calendar example (selleck)

multiple selection mode

enabled date range

Refactor Basic Calendar Navigator as plugin

sprint tracking ticket

Known issue: Chrome 12 -webkit-transform bug causes axis label positioning to fail

The Chrome 12 bug listed below causes issue with axis label positioning in YUI3 Charts.
http://code.google.com/p/chromium/issues/detail?id=85506

The defect can be seen here:
http://developer.yahoo.com/yui/3/examples/charts/charts-globalstyles.html

According to the google ticket, the bug fix will go into Chrome 13.

Need to explore possibility of a fix that will not break with future releases.

Charts - add more unit tests

Charts need more unit tests. Commit to 1.5 days of additional unit test coverage.

Charts - Readme/History file breakout

Need to break out history and read me files to new format.

Charts - convert api docs

Need to generate api docs in the new format.

TYPE_JSON should check for nested paths, then for full key string

For a JSON response that has flat records with keys that include dots, spaces, etc, DataSource.parseJSONData should first try to follow the parsed path to the nested data, but if it doesn't find it, to try again with unparsed key

ds.responseSchema.fields = [ { key: 'foo.bar' } ]

should look in
parsedJSON.records[0].foo.bar

then fail over to
parsedJSON.records[0]["foo.bar"]

This should alleviate a great deal of confusion wrt TYPE_JSON's special treatment of keys. If the order were reversed (unparsed key, parsed key), this could be used for TYPE_JSARRAY as well and maintain the current behavior and performance.

DataSchema API docs don't mention schema config fields anywhere

The API docs for DataSchema don't discuss any of the options for a schema anywhere, leading to some banging my head against a wall until I found them on http://developer.yahoo.com/yui/3/dataschema/

I could see the "meta" object in the response, but it was empty & I knew that couldn't be right so I dug into the code. Had the API docs had this information, I would've saved a good chunk of wasted time.

DS.IO + DataSourceJSONSchema with only metaFields fails

This is regression from 3.3.0 behavior.

I've set up a IO Datasource with a DataSourceJSONSchema to call an API that doesn't return a list of objects. It seems a bit weird but through the use of the metaFields in the schema I was able to avoid having to wrap my results in an array & instead got what I wanted as a neat tidy package.

It's tough to tell exactly what changed in the github diff, but something about _parseResults changed & it now sets a .error property on the data it returns. issueCallback in DS.Local sees that error field & calls the failure callback.

https://github.com/yui/yui3/commit/df5496ef3926fd0ea425cd4aa5eb278364c98d5c#L9R150

Would provide more useful feedback if I was more familiar w/ this code, sorry.

Pollable should start immediately

The Pollable interface allows you to poll a datasource. However, it starts polling first time after the interval seconds have passed. However, what I would want, and I expect is a more common scenario, is that polling starts immediately.

Of course, it is possible to write a separate request to initiate an immediate request, but that just duplicates code.

Alternatively, I would request and optional attribute: startImmediately.

Add aliase callback=>on for library consistency

The object passed to sendRequest() should use

 

{    on: {

        success: successHandler,

        failure: failureHandler

    },

    context: thisObj

}

for consistency with IO and other APIs.

datasource-function shouldn't fire the 'data' event twice on error

If you have a Y.DataSource.Function instance, and you make a request where the success callback throws an error, the datasource returns data twice.

For example:

This is because of this section in datasource-function.js:

try {
response = fn(e.request, this, e);
this.fire("data", Y.mix({data:response}, e));
}
catch(error) {
e.error = error;
this.fire("data", e);
}

The 'data' event is fired a second time if there is a problem with the success callback.

Would it be possible to fire a different event on error? Returning the data twice makes it extremely hard to debug problems, because in my specific use case (Autocomplete), the initial data returned was wiped out by the subsequent data returned.

Datatable does not display fields with utf-8 keys

Fields having a key with special characters are not displayed

I put an illustration of my problem at: http://www.nsa.be/temp/datasource-keys.html
The objects to be displayed in the datatable have a key "Ã©mail", corresponding to the Column Ã�mail. When the data is loaded, you can see in the firebug console that the key Ã©mail is present, but the value passed to the formatter is undefined.

The encoding is fine as the value "deuxiÃ¨me proposition" is displayed correctly.

A workaround was suggested by Satyam on the forum (http://yuilibrary.com/forum/viewtopic.php?f=90&t=2761&p=9363#p9363 ).
I confirm this workaround is working, as illustrated here:
http://www.nsa.be/temp/datasource-keys-ok.html

datasource-cache uses cached callbacks

datasource-cache is caching the entire datasource request object, including callbacks. when a cache request is successful, the cached request object is incorrectly mixed with the current request and the callbacks from the cache are used.

the response event is fired with a Y.mix of the cache entry and the request as its payload. however, the mix looks to be in the incorrect order

this.get("host").fire("response", Y.mix(entry, e));

the only properties of entry we really care about at this point don't exist in e. Just flip the order of the Mix and you preserve the request parameters (callbacks, tId, cfg).

All events have e.type === "dataSourceLocal:request" for at least DS.Local

True for subscribers to "request", "data", and "response".

I suspect what is happening is that the stack processing or the facade generation is going sour for the nested event firing with queuable: true events.

The event payload is in tact, so this appears to be an issue with the type property specifically.

DataSource.sendRequest without any arguments broken in 3.4.0 HEAD

In 3.3.0 it was possible to call .sendRequest() on a DataSource & not provide any arguments.

Due to https://github.com/yui/yui3/commit/961be3fe6b0d7f5051fb2e6b8f65b4505f5e67bc#L6R366, that no longer works & instead throws an error when it tries to access object properties that don't exist. The code does default the argument to an empty object, but it happens AFTER those properties are accessed so it doesn't help.

DataTableScroll plugin errors with DataTable + DataSource

This is a filing for this forum post. It has a link to an example on jsfiddle which uses 3.3.0pr3 but the problem persists in main. In short, a DataTableScroll plugin doesn't seem to work with a DataTable which has DataTableDataSource also plugged in. It appears that in datatable-scroll-debug.js line 338 the widget assumes the data is already loaded.

There is also a possibility that this is user error on my part, but so far it appears I'm one of the few in IRC and the forums that is testing out the DataTable and I haven't received any suggestions on how I might approach this differently. If that is the case, may I request an example which uses the 2 plugins be included?

[bz 4284252] XHR data load breaks sorting

Calling load after setting up sorting breaks sorting.

cannot create DataTable with no caption

If you don't pass a caption into new Y.DataTable.Base(), an empty one is created anyway, which causes ugly padding above the table element.

[bz 4462510] Column attributes formatter and sortable are not mentioned in documentation

Column attributes formatter and sortable are not mentioned in documentation

DataTable recordset is dumping array values as undefined when 'dump' module is not loaded

Here is a simple example:

YUI({ filter: 'raw' }).use('datatable', function(Y) {
var dt = new Y.DataTable.Base({ columnset: ['color'], recordset: [ { color: [ 'red', 'blue' ] } ] }).render();
});

Looks like Y.dump when the module is not loaded is set to a NOOP function on yui.js, which returns undefined when DataTable tries to output the value on the cell trough the Y.Lang.substitute.

Remove CSS Shadow in examples: Image-background and Image-surround for Safari

example css needs -webkit-box-shadow:none;

Center button reset string stays on after use on touch interface

Fix so that it doesn't always display after clicking on touch devices.

Dial Smoke test fails - only during build process

It passes when the url is run onâ�¦
1. The XP IE 8 box near Allen
2. The clearafraid-dx build box
3. My XP FF, Chrome, Safari, IE6
It only fails when run as part of the Build. Reid demonstrated this.
Reid also demonstrated it is not an issue of window focus.

Failure is on "test mousedown on three oclock"
Asserting on Dial Value not XY of handle

Example for Delegate Drag and Drop has wrong parameters

In this example pages for DD Delegate the "container" init attribute reads "cont" - and it's wrong.



YUI().use('dd-delegate', 'dd-drop-plugin', function(Y) {

    var del = new Y.DD.Delegate({

        cont: '#demo', // agh... it should be container: '#demo'

        nodes: 'li'

    });

});

http://developer.yahoo.com/yui/3/examples/dd/delegate-drop.html
http://developer.yahoo.com/yui/3/examples/dd/delegate-plugins.html

The actual examples are correct and working, though.

Using DDNodeScroll with DDDelegate causes JS error in Safari

In Safari, if you try and pass the DDNodeScroll plugin to the DD.Delegate dragConfig, when you try and drag there are numerous JS errors. It looks like it has to do with this line: https://github.com/yui/yui3/blob/master/src/dd/js/delegate.js#L117 When you try and copy the dragConfig via Y.clone, Safari seems to mangle the value of "node" in the DDNodeScroll constructor if it is an actual Y.Node instance.

Incorrect documentation

The code here: http://developer.yahoo.com/yui/3/examples/dd/photo-browser.html for Y.delegate is incorrect.. It reads:

Y.delegate(function(e) {
if (!e.shiftKey) {
//No shift key - remove all selected images
wrapper.all('img.selected').removeClass('selected');
}
//Check if the target is an image and select it.
if (e.target.test('#yui-main .yui-g ul li img')) {
e.target.addClass('selected');
}
}, document, 'mouseup', '*');
//Listen for all clicks on the '#photoList li' selector
Y.delegate(function(e) {
//Prevent the click
e.halt();
//Remove all the selected items
e.currentTarget.get('parentNode').all('li.selected').removeClass('selected');
//Add the selected class to the one that one clicked
e.currentTarget.addClass('selected');
//The "All Photos" link was clicked
if (e.currentTarget.hasClass('all')) {
//Remove all the hidden classes
wrapper.all('li').removeClass('hidden');
} else {
//Another "album" was clicked, get its id
var c = e.currentTarget.get('id');
//Hide all items by adding the hidden class
wrapper.all('li').addClass('hidden');
//Now, find all the items with the class name the same as the album id
//and remove the hidden class
wrapper.all('li.' + c).removeClass('hidden');
}
}, document, 'click', '#photoList li');

it should be:

it should read:

Y.delegate('mouseup', function(e) {
if (!e.shiftKey) {
//No shift key - remove all selected images
wrapper.all('img.selected').removeClass('selected');
}
//Check if the target is an image and select it.
if (e.target.test('#yui-main .yui-g ul li img')) {
e.target.addClass('selected');
}
}, document, '*');
//Listen for all clicks on the '#photoList li' selector
Y.delegate('click', function(e) {
//Prevent the click
e.halt();
//Remove all the selected items
e.currentTarget.get('parentNode').all('li.selected').removeClass('selected');
//Add the selected class to the one that one clicked
e.currentTarget.addClass('selected');
//The "All Photos" link was clicked
if (e.currentTarget.hasClass('all')) {
//Remove all the hidden classes
wrapper.all('li').removeClass('hidden');
} else {
//Another "album" was clicked, get its id
var c = e.currentTarget.get('id');
//Hide all items by adding the hidden class
wrapper.all('li').addClass('hidden');
//Now, find all the items with the class name the same as the album id
//and remove the hidden class
wrapper.all('li.' + c).removeClass('hidden');
}
}, document, '#photoList li');

It is set correctly on the actual example page:

http://developer.yahoo.com/yui/3/examples/dd/photo-browser_source.html

DragDrop enhancement - ability to use dd with non-node objects.

Below is how I would like to use dd with a shape instance.
http://www.trippbridges.com/yui3/src/graphics/tests/manual/drag.html

Below is link to a version using a plugin that does a capability check instead of node check:
http://www.trippbridges.com/yui3/src/graphics/tests/manual/shapedrag.html

The code is here:
http://www.trippbridges.com/yui3/build/shape-drag/shape-drag.js

Editor isn't retrieving saved content in YUI 2-in-3 example

See this example:
http://devel.yuilibrary.com/yui/docs/yui/yui-gallery2.html

The editor allows you to save the content, but when the page is refreshed, the saved content is not retrieved.

Does it make sense for Y.Escape.html to convert non-strings to strings internally - as a convenience.

Otherwise you end up having to do:

var unknownvalue = ...; // lets say it's the number 1234
var sanitized = Y.Lang.isString(unknownvalue) ? Y.Escape.html(unknownvalue) : unknownvalue;

Y.Escape.html(unknownvalue + ""); // ugly

Y.on( 'available' ) docs need to be clear about wanting a selector



Y.one( 'available', function () {}, Y.one( '#some input' );

Throws JS errors: "s.replace is not a function" and "J.match is not a function". In both cases, the object (s/J) is the DOM Node instead of a Y.Node object.

See the URL I provided for a simple test case. The expected outcome is that the input gets focus when it's available.
I tried this with 3.1.1 and 3.1.0 too, with the same result.

A propos: What I'm trying to do is focus() on an input I create (and append) through Y.Node. The code in the example shouldn't throw errors, but maybe I'm going about this the wrong way anyway?

[bz 4277391] "gesturemove" event is fired after mouse button is up, if there is no "gesturemoveend" event handler

see attachment to bugzilla ticket for code sample that reproduces bug

description:

As attached "pageto reproduce this bug" html shows, "gesturemove" event is keeping fired after
mouse is up.

However, if an event handler is attached to "gesturemoveend" event, no "gesturemove" is fired after
mouse button is up.

This is counter-intuitive and must be a defect.

This can be reproduced in both 3.2.0 and 3.3.0.

Flick needs to wait till first move before logging start time, to get accurate velocity

Right now, flick uses mousedown/touchstart as the startTime for velocity calculations (distance/time). If the user mousesdown/touchstarts but doesn't move their finger for say 5s, then the velocity calculations are way off. We need to use the first touchmove, mousemove as the startTime for velocity calculations

Event.button and Event.which don't work in IEs 6 to 8

This is weird. Consider the following code:

Y.on('click', function(e) { alert(e.button); }, 'body' ); // Alerts &acirc;��1&acirc;�� in Chrome/Firefox and &acirc;��undefined&acirc;�� in IEs 6 to 8

According to YUI3 docs, this â��.buttonâ�� property should be supported by the YUI event object and should return â��1â�� for left click. It doesnâ��t seem to work in IE.

The really odd thing is that the *native* event objects seem to support â��buttonâ�� in all browsers â�� so YUI actually appears to be *breaking* functionality that was working perfectly fine:



// e._event passes through to the native browser event object

Y.on('click', function(e) { alert(e._event.button); }, 'body' ); // Alerts &acirc;��0&acirc;�� in Chrome/Firefox/IE6-9

Regards, Robin.

Body keydown events are broken

This code:



function keydown (ev) {

    switch (ev.keyCode) {

        case 32: // space bar

        case 34: // page down

        case 39: // right

        case 40: // down

            alert("next");

            break;

        case 33: // page up

        case 37: // left

        case 38: // up

            alert("previous");

            break;

    }

}

Y.on("key", keydown, "body", "down:");

...worked in 3.3.0, but fails to work as of 3.4.0PR1.

For a complete example, see Test Information below.

[bz 4626707] YUI 3 FocusManager: API docs should mention that default value of circular is true

FocusManager: API docs should mention that default value of circular is true

Run CSS Lint on yui3-skin-night files and clean up

http://csslint.net/

Skin Night: left pointing button doesn't zoom gracefully with browser zoom

When you use browser zoom, the left pointing button in the night skin doesn't align with the button.
Button is not formally adopted in 3.4.0 yet.

Design support for YUI website

Provide any designs, sketches, or discussions of UI for the website.

Port configurator

Graphics - additional unit tests

Graphics have some unit tests but need a more comprehensive set. Commit to 1.5 days of additional unit tests for sprint 3.

Graphics - user guide and examples

Need user guide and examples for Graphics.

Graphics - Readme/History file breakout

Need to break out history and read me files to new format.

Graphics - API Docs

Need to generate api docs in the new format.

Bad check when tacking on non-highlighted remainder of haystack

Off-by-one check for remaining haystack chars causes unexpected behavior when there is only one char remaining on the haystack.

Current behavior:
Y.Highlight.allFold('dsw', 'ds')
=> <b class="yui3-highlight">ds</b>

Expected behavior:
Y.Highlight.allFold('dsw', 'ds')
=> <b class="yui3-highlight">ds</b>w

Add ability to get config of a transaction

It would be nice if you were able to get the config from an IO transaction either by transaction id or provide the config in the Event payload.

I'm writing a Console add-on for debugging and it would be nice to listen to IO events globally and display that info in the Debugger.

Can IO use event-custom and event facades?

It would simplify the story of the most common methods/objects/code paths in YUI 3 if IO passed event facades to the callbacks. This way DOM event subscribers and Attribute change event subscribers and IO event subscribers would all share the signature callback(e, argN...) where e has a type and relevant properties.

I know this would not be backward compatible if just changed in isolation, so I'm curious if there is a way this could be implemented via a configuration or alternate method?

XDR module breakdown into IE and Flash sub-modules

task added for sprint tracking

IO instance and external transports unit tests

task added for sprint tracking

Migrate examples to Selleck

Convert all IO examples to Selleck.

IO - sending data together with form serialization in GET mode doubles up data params

IO kind of assumes that data will only be sent together with form serialized data in POST mode, as a result sending extra data together with form serialization in GET mode doubles up the params passed in data .

Now, it looks like you assume that sending form data in GET mode shouldn't be combined with also sending extra data, I might be overlooking something, however I don't see the point.

EvalError when using non-native JSON

On browsers that don't have a native JSON implementation the code branch in Y.JSON makes use of _eval, which is defined as _eval = fromGlobal('eval').

This results in an EvalError - "function eval must be called directly, and not by way of a function of another name"

This was discovered in BlackBerry Browser (version 5, simulator)

jsonp don't has a charset config

Look at jsonp code,there isn't a charset config,it's utf-8 by default
But my page's charset is GBK in china

Can u add a charset config in jsonp?



YUI.Env.JSONP[proxy] = wrap(config.on.success);

Y.Get.script(url, {

            onFailure: wrap(config.on.failure),

            onTimeout: wrap(config.on.timeout, true),

            timeout  : config.timeout

        });



YUI.Env.JSONP[proxy] = wrap(config.on.success);

Y.Get.script(url,{

            onFailure: wrap(config.on.failure),

            onTimeout: wrap(config.on.timeout, true),

            timeout  : config.timeout,

            charset : config.charset

        });

Add combo-module-separator property to YUI Loader to support combo-handling strategies other than query-strings.

The Loader code and YUI config are very flexible, but force you to use a query-string combo-handling strategy.

A complement to: comboBase and root, would be a comboModuleSeperator. By default this property should be set to "&", but having this property available in the YUI instance configuration would allow the YUI Loader to dynamically construct combo-URIs which could use varying approaches, and not restricted to query-string params.

An example could be to use path-matrix-params:

comboBase: "/yui3/combo;"
root : "3.0.0.build974/"
comboModuleSeperator : ";"

Would result in Loader dynamic generated combo URIs like:

/yui3/combo;3.0.0.build974/attribute/attribute-min.js;3.0.0.build974/base/base-min.js

I feel pulling out the "&" into a configuration property would give some desirable added flexibility to dynamic combo-handling.

Should loader have a public method to generate paths from the sorted module list?

It's something we do manually in the configurator:

if(current.combo) {
var filter = loader.FILTER_DEFS[current.filter.toUpperCase()];
var filteredPath = (filter) ? m.path.replace(new RegExp(filter.searchExp), filter.replaceStr) : m.path;
comboJsUrl.push(loader.root + filteredPath);
} else {
url = m.fullpath || loader._url(m.path);
jsout.push('<script type="text/javascript" src="' + url + '"></scr' + 'ipt>');
}

And others who use Loader as an offline dependency calculation API need to do the same.

Loader default path incorrect, when using it as an offline dependency calculation service

I end up having to do this to get Loader to set a base path correctly in configurator...

var loader = new Y.Loader({...});

// Set a base path if the user provides one (current.base is the base path the user enters in the text box), else default to yui.yahooapis.com/[version]/build/
if (current.base != "") {
loader.base = current.base;
} else {
loader.base = Y.Env.meta.base + Y.Env.meta.root;
}

loader.calculate();

a) The default base path doesn't seem to be accurate - hence the workaround in the "else" block (see #2528905).
b) Should loader pick up the base path from whatever is set for the YUI instance it's inside? Without the code in the "if" block above, it doesn't seem to.

datatype-xml missing

(12:06:32 PM) brian_endpoint: thoughts on the lack of yui3/build/datatype-xml/datatype-xml.js in HEAD?
(12:07:46 PM) brian_endpoint: perhaps it got split into datatype-xml-format and datatype-xml-parse
(12:08:05 PM) brian_endpoint: or it is one of the breakages from removing rollups?
(12:08:46 PM) rgrove: If datatype-xml was a rollup, then it has been split out into only its non-rollup components (intentionally, not a bug).
(12:09:03 PM) Tivac: looks ok... https://github.com/yui/yui3/blob/master/build/loader/loader.js#L3173
(12:09:13 PM) brian_endpoint: seems like something is trying to load it, I'm not (directly)
(12:09:41 PM) rgrove: Ah. If it's Loader that's trying to load it, then that could be a bug.
(12:09:52 PM) brian_endpoint: I'm assuming it is loader
(12:10:12 PM) Tivac: io-xdr & datatype load datatype-xml
(12:10:22 PM) Tivac: according to built loader.js
(12:11:00 PM) rgrove: Well, they require the rollup, but Loader should then load the individual rollup files rather than trying to load datatype-xml.js.
(12:11:27 PM) rgrove: s/rollup files/module files/
(12:11:31 PM) Tivac: yes
(12:12:20 PM) brian_endpoint: umm...
(12:12:26 PM) brian_endpoint: not sure where that leaves me :-)
(12:13:06 PM) brian_endpoint: /* This file is auto-generated by src/loader/meta_join.py */
(12:13:12 PM) brian_endpoint: meta_join.py doesn't seem to exist :-)
(12:13:39 PM) brian_endpoint: I suspect that is just an out of date comment, but thought I'd point it out
(12:13:46 PM) Tew: it's meta
(12:17:15 PM) brian_endpoint: looks like 8ebea46a72db798f14907a09fa8ce99422f01809 (from ls_n) is the break point
(12:17:26 PM) brian_endpoint: "Purge rollup files and update build.xml"
(12:17:27 PM) brian_endpoint: :-)
(12:17:53 PM) ls_n: That was just for my modules. Something broke after that?
(12:18:46 PM) rgrove: Well, the old datatype-xml.js rollup file would have existed until that point. That doesn't mean Luke broke it, it just means that when Luke removed the old built files, the breakage was revealed.
(12:18:52 PM) brian_endpoint: if I switch from 94dd -> 83be I start getting a missing yui3/build/datatype-xml/datatype-xml.js
(12:19:14 PM) ls_n: Sure. The file is gone.
(12:19:25 PM) brian_endpoint: okay...so...?
(12:19:26 PM) brian_endpoint: :-)
(12:19:40 PM) ***brian_endpoint is just a lowly user trying to make his way in the world :-)
(12:19:47 PM) ls_n: You have to refer to a combo of datatype-xml-format/...js&datatype-xml-parse/...js
(12:20:10 PM) ls_n: or use('datatype-xml') should still work to fetch the two modules
(12:20:10 PM) brian_endpoint: well... or someone does
(12:20:35 PM) ls_n: or load the scripts individually, I suppose. Not that that's a recommended solution.
(12:21:04 PM) brian_endpoint: I'm not directly using it, it is being pulled in by loader
(12:21:12 PM) rgrove: Again, this sounds like a Loader bug if datatype-xml isn't automatically being resolved into its submodule files.
(12:21:19 PM) ls_n: yep
(12:21:25 PM) brian_endpoint: okay, want for me to file it?
(12:21:27 PM) ls_n: should be in the aliases
(12:24:06 PM) rgrove: brian_endpoint: Yes please.
(12:24:40 PM) brian_endpoint: boy this rollup change is madness :-)
(12:25:14 PM) Tivac: confirm
(12:25:31 PM) ls_n: It simplifies the system, but maybe we're learning that it is a mistake...

Aliases: _explodeRollups issue with datatype

We came across this with the Charts component. It had a requires: ["datatype"] entry in it's meta-data and the end result was that loader attempted to pull in the physical build/datatype-number/datatype-number.js file, as opposed to breaking up the alias into it's respective pieces [ datatype-number-parse.js and datatype-number-format.js ]

Have a feeling this may be unique to the fact that datatype is a rollup of rollups, or that datatype-date has supersedes meta-data, because in general aliases instead of rollups are working fine.

To replicate it in it's simplest form, I just created a dummy component with the following meta-data in yui.js/loader.js:

"foo": {
requires:["datatype"]
}

And when using this component:

YUI({filter:"raw"}).use("foo", fn() { } )

And stepping through code, I saw that in the loader.calculate steps...

This was the this.required list after the _explodeRollups() step:

foo: true
datatype-date: true
datatype-date-format: true
datatype-date-parse: true
datatype-number: true
datatype-number-format: true
datatype-number-parse: true
datatype-xml: true
datatype-xml-format: true
datatype-xml-parse: true
event-custom-base: true
event-custom-complex: true
intl: true
intl-base: true
lang/datatype-date-format_en-US: true
oop: true
yui-base: true

And then after the _reduce() step, it was whittled down to this:

foo: true
datatype-date: true
datatype-date-parse: true
datatype-number: true
datatype-xml: true
event-custom-base: true
event-custom-complex: true
intl: true
intl-base: true
lang/datatype-date-format_en-US: true
oop: true
yui-base: true

Which is wrong (it left the datatype-number rollup in the list, and removed the leaf modules, instead of the other way around). This is because datatype-number.supercedes was [datatype-number-parse, datatype-number-format].

Other interesting points:

a) I noticed _reduce was called I think 3 times for this test. Wondering if this is expected? I would have imagined that it only gets called once or maybe twice (once for the core modules and once for the use'd modules maybe)
b) If I take the same module meta-data and add it to the config instead of directly to loader meta-data is works fine:

// Works fine. It's only when I move this to loader.js/yui.js that it fails.
YUI({
modules: {
foo: {
requires:["datatype"],
fullpath: "foo.js"
}
},
filter:"raw"
}).use("foo", function(Y) {

});

c) NOTE: Charts is being changed to require datatype-number and datatype-date (as opposed to the mega datatype rollup) because that's all it really needed. So you may not be able repro the bug with the latest charts. You'll probably need to set up a dummy component as mentioned above (I actually just hacked the charts meta-data).

meta_join.py or something in the loader build process looks like it may be ignoring paths

For example, yui3/src/cssgrids/meta.json has:

[master] $ cat cssgrids.json
{

"cssgrids": {
"path": "cssgrids/grids-min.css",
"optional": [
"cssreset",
"cssfonts"
],
"type": "css"
}
}

But what ends up in loader-yui3.js is:

"cssgrids": {
"optional": [
"cssreset",
"cssfonts"
],
"type": "css"
},

I understand this was probably done consciously for the YUI out-of-the-box modules because we're flattening our dir structure, but longer term we should probably clean up the meta-data files which still have paths, and revert meta_join.py to actually honor paths if they exist. That way if someone wants to use meta_join with their own custom meta-data they can, and it's not a hidden assumption built into the tool.

Basic RLS example failure: Uncaught TypeError: Object [object Object] has no method 'rls_locals'

When using RLS with YLS 0.4.0pre, the following error occurs:



    Uncaught TypeError: Object [object Object] has no method 'rls_locals'

This appears to occur because the rls module isn't bound to Y.

Conditional modules triggered off of rollups failing to trigger

With the removal of rollups in the loader config & their replacement with aliases outside of the actual loader config any conditional modules that were set to trigger on a rollup are now never triggered.

It's an easy fix (datatable => datatable-base) but tripped me up for a bit and definitely seems like a bug.

Slider one-off skins not being loaded

Slider has one-off skins, like sam-dark and capsule that are being built into /build/slider-base/assets/skins/.

The following code does not load these skin files:

YUI({
skin: {
overrides: {
slider: [
'sam', // The default skin
'sam-dark', // Suited for dark backgrounds

'capsule', // You only need to include one skin
'capsule-dark', // in the overrides section unless you
// are using multiple skins on the same page
'round',
'round-dark',

'audio-light',
'audio'
]
}
}
}).use('slider', function ( Y ) {

Test pages:
yui3/src/slider/tests/manual/skins.html
http://stage.yuilibrary.com/yui/docs/slider/slider-skin.html

Convert Examples

ticket added for sprint tracking

memory leak in Node.replace, when Node.create returns a fragment

when given a string argument, Node.replace calls Node.create. If Node.create returns a fragment, Node.replace should destroy the fragment, but it doesn't, so the fragment Node stays in _instances.

Please see http://jsfiddle.net/TTNHx/1/ for an example.

Overlay not completely being destroyed on destroy()

When destroying an overlay in 3.4 and then scrolling the page an error is thrown every time the page detects the scroll.

The errors aren't present in 3.3

Y.Plugin.Modal should fire events before and after mask is shown

Fire "maskShow" and "maskHide" methods as appropriate

Results table added to bottom of the page

Clicking on the 'view created tables' a second time adds another results table to the bottom of the page.

RecordSet.getValuesByKey incorrect documentation

The documentation comments for the getValuesByKey method on Recordset contain

@param index {Number} (optional) Index at which the required record resides

There is no optional index parameter. There is a string parameter named key which is required.

ScrollView pagination should be supported along both the X and Y axis.

ScrollView paginator currently only works along the Y or horizontal axis (as part of the original implementation). It should be enhanced to work along both.

[bz 4610978] Dragging finger outside the viewport (and onto browser chrome) results in erratic scroll behavior

Prerequisite:
You need a scroll view with other content placed under it. Can be reproduced with the example in the YUI docs: http://developer.yahoo.com/yui/3/examples/scrollview/scrollview-base_source.html

Steps on iPhone:
Scroll the view down, so you have enough room to scroll it up
Tap the scroll view and drag finger down to start scrolling the view up
Continue drag your finger down (without lifting it), drag down out of the scroll view and over the content under it all
the way down over the browsers navbar (and/or outside the screen).
Lift finger and put it down on the scroll view again to continue scroll up

Result:

In step 4 the scroll view will jump back to where you started the scrolling in step 2. (So you can continuously scroll
the same area over and over again, without reaching the top of the view)

Other info:

* Weâ��ve tested this on iPhone (iOS version 4.0.2, 4.2, and 4.3)
* It is reproducible on iPhone (and iPhone simulator) with iOS version 4.0.2.
* It is not reproducible on iOS 4.2, and 4.3.
* We havenâ��t looked into the issue further. But it looks like maybe the touchend event is not fired on iOS 4.0.2
when you drag the finger outside of the viewport/screen and that could cause this behavior?

ScrollView-List Plugin

Currently ScrollView is content agnostic. It doesn't ship with CSS which references it's content.

Jeff needs a ScrollView-List plugin, which can be plugged into ScrollView-Base for plugins which have LI content so we can provide a stylized out of the box experience for ScrollView with LI content. All the plugin will likely do is add classes to the first LI descendents (or something more intelligent) which can be used to key rules off of. This is mainly to support IE6/7, where we can't use >.

cross-browser native Transitions

It looks like -webkit-transition is hard-coded in the code. FF4 supports transitions via -moz-transition. Please add support for that:

https://developer.mozilla.org/en/CSS/CSS_transitions

conditionally use transition-timer

use('transition') should only load transition-native when native support exists.

WidgetStdMod setStdModContent method is only valid for use after render

WidgetStdMod setStdModContent method is only valid for use after render, worth changing this to be available at any time of the Life-cycle?

potential memory leaks from Nodes after widget destruction.

Since YUI3 Node caches all Node instances (in its _instances module variable), any obsolete Node must be properly destroyed by calling node.destroy() or it will remain in _instances and memory will leak. Widget's destructor attempts to do so:



if (contentBox) { // Just to be safe because it's a last minute change. Really shouldn't be required.

    contentBox.remove(TRUE);

  }

  boundingBox.remove(TRUE);

However, remove(true) will call destroy only on the node it was called on (in this case, contentBox and boundingBox), but not on any other descendant nodes.

The assumption, however, seems to be that Widget will properly destroy all its nested nodes. For example, here is the destructor implementation from the spinner example (http://developer.yahoo.com/yui/3/examples/widget/widget-extend.html):



/*

 * destructor is part of the lifecycle introduced by 

 * the Widget class. It is invoked during destruction,

 * and can be used to clean up instance specific state.

 * 

 * The spinner cleans up any node references it's holding

 * onto. The Widget classes destructor will purge the 

 * widget's bounding box of event listeners, so spinner 

 * only needs to clean up listeners it attaches outside of 

 * the bounding box.

 */

destructor : function() {

    this._documentMouseUpHandle.detach();

 this.inputNode = null;

    this.incrementNode = null;

    this.decrementNode = null;

}

Note that .destroy() is not called on any nodes (and .remove(true) on contentBox/boundingBox will not call destroy on them either) - after this widget is destroyed, they will remain in _instances, resulting in a memory leak.

In this case, the Nodes were created explicitly through YUI (making it more apparent that a Y.Node was created but not destroyed), but a Node can also be created implicitly from any DOM element, for example if there is a subscription to the "mousemove" event, any DOM element that the mouse moves over will get a Y.Node if it didn't already have one (so the Node can be provided to the event listener). Or, do something like Y.all('*').each(function() {}); and now every DOM element has a Node in _instances. Because it is so easy to end up with cached Nodes, proper node cleanup seems to be critically important.

I don't know whether the Widget destructor should change to properly clean up its entire node hierarchy or the docs and examples should change to stress the importance or proper manual node cleanup. In our application we tried patching the Widget destructor to perform deep node cleanup, and it significantly reduced memory leaks.

You might also want to look at http://yuilibrary.com/projects/yui3/ticket/2530415, which gave rise to this ticket.

Abstract out events on which WidgetPositionAlign re-syncs

Currently, these events are hardcoded as "windowresize" and "scroll" - should be more generalized, maybe as an ATTR which takes an array of event strings.

Migrate Docs to Selleck

Attribute Examples
Intl Examples
Overlay Examples
Plugin Examples
ScrollView Examples
Widget Examples
Gesture Examples

(~24 examples)

Attribute Landing
Base Landing
Configurator
Intl Landing
Overlay Landing
Plugin Landing
ScrollView Landing
Widget Landing

WidgetStdMod has Header/Footer button support

Have the ability to add buttons to the header or footer of WidgetStdMod

Create Y.Panel

Panel = All extensions used by Overlay + Y.Plugin.Modal + Y.Plugin.Drag + Y.Plugin.Resize + Header/Footer button support

Re-write WidgetModality and WidgetAutohide as extensions instead of plugins

It makes the API much nicer, will be needed in Y.Panel

[Pull Request] - Fixed bug in yql module where options ('opts') were not actually being used. This prevented use YQL calls over SSL, which some tables require.

Discovered a bug today while trying to make a YQL request over HTTPS. Fixed.

Y.later crashes when passing native functions in IE7

Y.later(when, o, fn, data, periodic) attempts to execute "fn.call(o)", but methods on DOM objects in IE7 don't have a .call property and it crashes.

ie:

// focus this input box
Y.later(100, my_input_box, my_input_box.focus);

RLS needs to support more of the Loader config options..

fetchCSS and ignore should also be supported by RLS..

fetchCSS can be handled in the $rls callback, just check for the config and ignore the request..

ignore will require a new parameter on the url..