It is easy to exit the list at any given point by using the toolbar
button or pressing enter twice while at the end of the list.
Change-Id: Ie34300fc13a4548e283348ef58c88ce420fbe0c1
ce.GeneratedContentNode had an interesting bug where it called .append()
directly on DOM elements stored in the store. They weren't cloned,
which meant the previous rendering of the same node would just disappear,
and they also weren't adopted into the correct document which would
probably have caused other issues as well.
Properly clone and transplant the nodes from the store before attaching
them to the DOM.
Change-Id: I423db85cb7c3851a9bf68de03c72aa22994d9474
Because scroll events cannot be canceled or prevented from bubbling
up, the only way to prevent the scrolling is to cancel the actual key
and mouse wheel events.
PhantomJs -= 2
Change-Id: I540738459181c37d11caf5db07345703e7000ef9
This is mainly used for testing blacklisted browsers, so we just need
one parameter ('vewhitelist') which sets whitelist to true and
blacklist to false.
Change-Id: I38771ba7a49657d67c4d94ace1f30b5e41904df6
Objective:
* Make all text inputs have the ability to be "pending", not just ones
with a special mixin
Changes:
* Delete pending input widget
* Move pending input widget implementation to text input widget
* Update all uses of pending input widget
* Make pending image a reusable "texture"
* Update styles of text input widget for pending state
* Get rid of checking for mixin since all text inputs can be pending now
Bonus:
* Get rid of unused images, including .psd and .ai files
* Add transparency texture
* Fix input widget not using documented config value
* Fix documentation in select widget (lies!)
Change-Id: Ib46ab01dc39d706e5c25fd473dee0edce51b7e44
Browser history state:
* Listen for popstate, if veaction is different, act on it
(by calling activate or deactivate).
* When switching to edit mode, push veaction=edit.
When switching to view mode, push non-veaction.
Except if the switch was made by the popstate listener in
which case we don't want to push again.
Clean up:
* Use document.location.href instead of odd location.toString
Don't pass it to mw.Uri from the original one since it
takes the current url by default.
Bug: 43844
Change-Id: I5c345c1526852a84b04436955845bf1b56d94402
Merge methods:
* transformPage: transformSkinTabs + hideSiteNotice
* restorePage: restoreSkinTabs + restoreSiteNotice
Callers now consistently do both (e.g. #tearDownSurface only
called #restoreSkinTabs). This is also in preparation for
bug 43844.
Removed call from #tearDownSurface as that method is only
called from #deactivate which already calls #restorePage.
Clean up support check:
* Rename supportsStrictMode to supportsES5. Though the feature
test asserts strict mode, we use it to check support for
ES5. VisualEditor doesn't particularily care about strict
mode itself.
* Reverse support if-statemement to return early instead of
wrapping everything that follows. This makes it easier to
see that we intent to abort if the current environment is
known to be problematic. Also easier for code later on
by not accidentally falling outside this block
* Follows-up aaa5ad254b.
Change-Id: Ia4b949d9c066a3f7b07217aa3d51de9908734e85
This makes the 'category' section of the meta-data panel less lonely, and
prompts us to actually do this at some point quite soon (or encourage others
so to do).
Bug: 48814
Change-Id: I5b8fdfb78a2117839277a683db47fe97107d87b0
The update() method wiped out the entire contents of the node, including
the shield that makes ce.ProtectedNode work. Fixed by only wiping out and
reconstructing the contents of the <ol>
Change-Id: Ib2978a72939884be67964ce8a3d89a570f70bfa3
The previous fix didn't really work, because the notices are expected to
be an array elsewhere too. Better to just convert it to an array on the
spot. getObjectValues works with any kind of object (including an array)
so it's safe to use with either data type.
Change-Id: I03237b8624a0b980e5f70d54d55c662ffa460373
The easy part is getting the correct numbers from the InternalList
and generating the ordered list HTML. The tricky part is connecting
up the events to make sure the renumberings/list generations are
triggered when required.
InternalList can emit an update event on document transaction, which
triggers the renumbering/relisting if any references have been
added or deleted during that transaction.
ve.ce.MWReferenceListNode also listens to changes on the
InternalListNode (i.e. changes to the contents of the references)
and always triggers a rebuild.
Change-Id: I1b48ef5240e433c3b314259aa68cde13841ea98b
We now have three stages:
1. Browser feature tests. Dies silently if any fail.
2. Browser blacklist. Dies silently if match found.
3. Browser whitelist. Shows warning if no match found.
Previously we were treating the remotely generated
edit notices as if they were in an object when
in fact they were in an array - the code has been
fixed to reflect that fact.
As locally generated notices will typically require
parsed messages, we've also moved the notice rendering
to after onReady is fired.
Updated jquery.client to latest master from MediaWiki core
(needed for proper detection of Iceweasel, Android and Safari)
Bug: 38128
Change-Id: Idc5f4a23a2709264d869a91d00873c4e187bc470
Follows-up 2e76271 and 231a50f which made manual changes to
the static-loading files without updating makeStaticLoader
(thus causing the load order and if-statement to be out of
sync between different index.php files).
Updating makeStaticLoader to include those changes and applying
it to the other index.php file.
Change-Id: I9bbe97d85f663b1cffeb384d52b5cc54e2f6601b
This has no influence on Jenkins but can be used locally to
easily run certain tools. Since we already had `.jshintrc` in
our repo it was already possible to easily run JSHint from
the command-line locally. Taking that as a base the following
are new features:
* `grunt csslint`: Runs CSSLint on all css files
* `grunt qunit`: Runs QUnit (standalone) tests in PhantomJS
* `grunt test`: Runs jshint/csslint/qunit
* `grunt watch`: Runs the "test" command automatically whenever
a file is changed. You can keep this in the background so
whenever you save a file in your editor (e.g. Sublime Text)
it'll run the tests and if there is a failure, it'll throw a
bash error code causing your Terminal application to beep you
in whatever way your operating system does so (e.g. for
Mac OS X a red badge + jumping icon in the Dock). It will
continue to run in the background even after a failure so no
need to re-start watch after a failure.
* `grunt`: Runs the default task, which is 'test'.
Previously to use `jshint .` you had to:
* One-time install:
* install package -- nodejs npm
* npm install -g jshint
* Usage:
* cd VisualEditor; jshint .
Now, for grunt:
* One-time install:
* install package -- nodejs npm
* npm install -g grunt-cli
* cd VisualEditor; npm install
* Usage:
* cd VisualEditor; grunt
Change-Id: I7a4fdf4b6bf3f00cef15dc3e2c81eceb595aec7c
Adding a fairly loose .csslintrc file so that our code
passes it.
The following options cause warnings in our code and have been
disabled for now:
* adjoining-classes
> Don't use adjoining classes.
> .ve-ui-widget-disabled.ve-ui-textInputWidget textarea:focus {
* box-sizing
> The box-sizing property isn't supported in IE6 and IE7.
> box-sizing: border-box;
* box-model
> Using width with border can sometimes make elements larger than you expect.
> border: solid 1px #ccc;
> Using width with padding can sometimes make elements larger than you expect.
> padding: 0 0.75em 0 0.75em;
> Using width with border-right can sometimes make elements larger than you expect.
> border-right: 1px solid #eee;
* fallback-colors
> Fallback background-color (hex or RGB) should precede RGBA background-color.
> background-color: rgba(104,171,255,0.1);
* important
> Use of !important
> position: relative !important;
* outline-none
> Outlines shouldn't be hidden unless other visual changes are made.
> .ve-ce-documentNode[contenteditable="true"]:focus {
* qualified-headings
> Heading (h1) should not be qualified.
> .ve-ce-branchNode h1:empty:before,
* universal-selector
> The universal selector (*) is known to be slow.
> .ve-ce-protectedNode * {
* unqualified-attributes
> Unqualified attribute selectors are known to be slow.
> .ve-ce-documentNode[contenteditable="true"]:focus {
Fixes made:
* modules/ve/ce/styles/ve.ce.Surface.css:
[L66:C2] margin can't be used with display: inline.
* modules/ve/ui/styles/ve.ui.css: Unknown @ rule: @-ms-keyframes
Internet Explorer < 10 doesn't support keyframes. IE10+ supports
the standard property. I don't think ms-keyframes ever existed
http://caniuse.com/css-animation. ms-transform did exist
http://caniuse.com/transform.
Change-Id: I728a48e489c079e1c94a506bb00c245de9551eb6
Due to the "es5: true" jshint option we enabled, these
warnings were surpressed. I've disabled the option since
we no longer require it. It was enabled in 07c86fc to fix
a bug with jshint. This bug has now been fixed.
Change-Id: I55b7d031eb5581af5f733f050cf2ea98dacb2af6
Also keep items in the order they appear in the document
and grouped by group and key.
Additions and removals are triggered by the new root/unroot events.
Change-Id: Ia3e90ccfdab88f352b89992b90554e5f03ff9952
Clean up of logic implemented during the template-sprint:
* Store spec inside the content model, directly associated
with the content-part. This allowed fixing the bug where
two spec-less template invocations overwrote each other's
made-up template data due to it using "target.wt" as key.
The opener now provides the fetcher with a "specId" which
is set to "part/<id>" for wt-generated template targets.
* Batching is now implemented inside the fetcher instead
of outside. This allows calling "getTemplateSpecs" inside
the loop with a dedicated callback for each spec to store
it in the content.parts[i] object passed by reference.
It also makes it easier to use by different code paths.
You call it as much as you like and it will queue up
naturally through javascript yielding and then make a batch
request. This is based on the pattern I used in MediaWiki core
for mw.loader#addEmbeddedCSS.
Follows-up e7af635, da679b7.
Change-Id: I4d7121229d060a96d927585c987a1a81a474b922
This is a hack, in the future Parsoid will have a template re-expansion
API that will produce the correct result.
Change-Id: I328c11330fb3db71c51882717d6b84099c9270d1
unicodejs.js:
* add splitClusters(text) and splitCharacters(text) methods
unicodejs.textstring.js:
* change internal representation from a char string to a list of grapheme
clusters
unicodejs.wordbreak.js:
* change getGroup to work on the first character of a grapheme cluster
ve.js:
* Use new unicodejs.splitClusters function
Bug: 48975
Change-Id: I202b98199d2780534d1e02519b72579ba796f08f
Changes:
ve.ui.Widget.css
* Add styles for decorated text input widgets and their icon elements
ve.ui.TextInputWidget.js
* Add icon option which adds an icon before input text
Change-Id: Ib48d795391cb5d110e7dc05658d51129792dfc33
Objectives:
* Break template page creation into it's own method
* Get rid of styling being applied in JavaScript
* Fix styling issues
Changes:
ve.ui.MWTemplateDialog.js
* Add method for adding a template page
* Replace css calls with addition of classes for styling
* Cleanup of append calls
ve.ui.Dialog.css
* Make inputs in the template dialog full width and a reasonable height
ve.ui.Widget.css
* Swap margins with padding in labels to prevent layout issues
* Prevent textareas from being resized in safari/chrome
Change-Id: I4030d8605aad865251ecd0aeb8cc72d333bed6a4
Objectives:
* Refactor template dialog to support loading template data for, and
editing multiple templates and interleaved content in a template node
* Update template node model to generate multi-template wikitext
Changes:
ve.dm.MWTemplateNode.js
* Rewrite getWikitext method to work with multi-template formats
ve.dm.MWTemplateDialog.js
* Retain information about the node and template calls
* Break AJAX handler into its own method
* Attach event handlers to inputs directly so template values are
edited directly on the fly
* Refactor page building to support multiple templates
* Add multi-template support for template data API call
* Add support for editing plain text content
Change-Id: I92ff8a9e186701a3f8da88def92a5b7dcb607897
Objectives:
* Fix compareObjects so it doesn't break when given arrays
* Remove compareArrays
* Rename compareObjects to compare and update callers of both methods
Changes:
ve.js
* Loosen check for whether to recurse so both arrays and objects qualify
* Remove compareArrays
* Rename compareObjects to compare
ve.dm.MWTemplateNode.js, ve.dm.AnnotationSet.js, ve.dm.Document.js,
ve.dm.Transaction.js
* Update uses of compare(Arrays|Objects) to use compare
Change-Id: I7d4f7ceb28c0389f0157b7598e291f21393b5b85
* In getDataElementOrSlice(), we were slicing one too far
* When encountering a closing for an internal node, don't traverse
up. Doing this caused weird bugs, like inserting text where it didn't
belong in some cases an exceptions in others, but these issues were
parly masked by the off-by-one error in the data slice.
Change-Id: Ieda9afa95b7c1953d09e391774350a9b4148c2fe
Objectives;
* Cleanup the addPage API
* Add optional index param to addPage method
Changes:
ve.ui.MWMetaDialog.js, ve.ui.MWTemplateDialog.js
* Update uses of addPage
ve.ui.PagedDialog.js
* Bundle existing optional arguments into object
* Add index option for inserting a specific index
Change-Id: Idcef4d0a52fb817c7d888990920b2c12224a3392
This will make sure the marker moves correctly (backwards/forward)
in RTL languages as well as LTR languages, judging from the wiki pageLanguage.
This can be a quickfix until the movement can be decided per the direction
of the specific element (span/paragraph/div) the marker is in.
Bug: 38546
Change-Id: Ic01e110a5e6094cd275327a2e8cea90c900f1bd1
Making sure resize handle events are bound and unbound on focus, and that elements are created in the right window.
Change-Id: Ie90bb82aa6c81c372d76278dab3665bd49bf573c
Trevor tweaked the pixel grid snapping on my previous commit
but it appears that introduced some errors as well. The horizontal
lines in underline-a and strikethrough-s no longer have integer
x-coordinates, and on underline-u the whole object is offset
by (0.166, 0.166).
Bug: 47780
Change-Id: I51f2605e7d8e2bab1d641f02244d5cd24f505676
When you typed after a ProtectedNode that was previously at the end of
a paragraph, then moved the mouse, you'd get a JS error.
What happened was the typing caused a new TextNode to be created after
the ProtectedNode, which caused the ParagraphNode to be rebuilt,
detaching the original ProtectedNode and creating a new one. The
detached ProtectedNode was still bound to mousemove on the body, and
when that event fired it would try to remove its phantoms from the
surface and fail because detached nodes can't get to the surface.
Change-Id: I9f38776f0267645b14d7b26e2a25007cf3be8ec7
The only reason it doesn't save correctly right now is because Parsoid
doesn't serialize reference edits correctly.
Change-Id: Ia0f272c07cc28ee829372eb848f23aec99eb92f0
Add merge() methods to IndexValueStore and InternalList, which merge
another store/list in to the current one and return a mapping
translating old indexes to new ones.
Also add functions that, given such a mapping, traverse a linear
model data array and remap store/list indexes using simple logic for
annotations and node type-specific functions for node attributes.
Change-Id: I1e90755ced1a87c190947c037cf151c4d17cf8b7
A document slice is a document built from a data slice of an existing
document. It's completely separate from the original document and has
its own store and internalList. The new document's data also contains
the entirety of the original document's internal list. It's possible
to create a document slice of data located inside the internal list,
in which case the resulting document will contain that data twice (one
mutable copy at the top level, and one immutable copy in the internal
list).
ve.dm.Document.js:
* Optionally take an internalList in the constructor. This allows us to
create a document with a clone of an existing internalList rather than
an empty one.
* Add edgeMetadata flag to getFullData()
ve.dm.IndexValueStore.js, ve.dm.InternalList.js:
* Make these classes cloneable
Change-Id: I93e06f764ace16aee9df941b07f8c2bff1a28e2b
Getting & setting the cursor is done with byte offsets
instead of data model offset (characters) so we need to
be able to convert between the two as well as splitting
characters.
TODO: The regex only works on surrogate pairs, not
yet combining accents.
fixupInsertion will combine a combining mark with the
character to its left it it can.
Bug: 48630
Change-Id: I8d936fb15d82f73cd45fac142c540a7950850d55
Add teardown call to surface destruction in mw target, and
teardown listener to resizeable node.
Bug: 48530
Change-Id: I807a0f32d3d1eb490456d887f7bf867bdb896df4
MW meta is close to being moved out of experimental, but language
editing is not in line for the beta release.
Bug: 48561
Change-Id: Ie87a3c7dde2f29c3898c08d4cade5c0ba2f937dd
Also remove the exception thrown when we try to add an non-existent
toolbar button, as it may just be experimental and not loaded.
Change-Id: I0a60421f45d7a3941c510defc60d1fbf9469e784
Right now the internal list is very much HTML-based, so it's not
populated at all when creating a document from data. So hack this in by
specifying what should be in the internal list as a property of the
example document's data array.
Change-Id: I51c7b7b4dcd9fd3333777c1287b7ba544887aa32
Prologue:
Farewell ve.Editor my good chap… Oh, hey there HTML frames - I didn't
see you there! In a world where iframes are outlaws, and symbols like
document and window are global, there were more than a few assumptions
about which document or window was being used. But fear not - for this
commit (probably) tracks them all down, leaving a trail of
iframe-compatible awesomeness in its wake. With the great ve.ui.Surface
now able to be used inside of iframes, let the reference editing
commence. But there, lurking in the darkness is a DM issue so fierce it
may take Roan and/or Ed up to 3 whole hours to sort it out.
Note to Roan and/or Ed:
Editing references seems to work fine, but when saving the page there
are "no changes" which is a reasonable indication to the contrary.
Objectives:
* Make it possible to have multiple surfaces be instantiated, get along
nicely, and be embedded inside of iframes if needed.
* Make reference content editable within a dialog
Approach:
* Move what's left of ve.Editor to ve.ui.Surface and essentially
obliterate all use of it
* Make even more stuff inherit from ve.Element (long live this.$$)
* Use the correct document or window anywhere it was being assumed to be
the top level one
* Resolve stacking order issues by removing the excessive use of z-index
and introducing global and local overlay elements for each editor
* Add a surface to the reference dialog, load up the reference contents
and save them back on apply
* Actually destroy what we create in ce and ui surfaces
* Add recursive frame offset calculation method to ve.Element
* Moved ve.ce.Surface's getSelectionRect method to the prototype
Bonus:
* Move ve.ce.DocumentNode.css contents to ve.ce.Node.css (not sure why it
was separate in the first place, but I'm likely the one to blame)
* Fix blatant lies in documentation
* Whitespace cleanup here and there
* Get rid of ve.ui.Window overlays - not used or needed
Change-Id: Iede83e7d24f7cb249b6ba3dc45d770445b862e08
The spec you link to appears to not correlate to the real world.
Firefox has a value of 0x2E (46 in decimal) store for this, and
both FF & Chrome report that value in my tests.
Change-Id: I4b0d4d27448587ca7381c640d6d3949402305656
* VisualEditor.hooks.php
** export URL to magnify clip as a configuration variable so thumbnail image can render correctly
* ve.ce.Surface.css
** add CSS styling to make image captions look the same way in edit mode as they look in the view mode
* ve.ce.ParagraphNode.js
** add CSS class ve-ce-generated-wrapper if it is a generated wrapper
* ve.ce.MWImageCaptionNode.js
** make image caption rendering match the view mode
Change-Id: I0cd1b25e8f8355e0500aabc90e7c4cdf591545f3
On wikis where certain "topicons" exist the toolbar showed a
weird blank space pushing the entire toolbar down and
ve-ui-toolbar-actions to the left.
Bug: 48734
Change-Id: Ic5f73ac1eb8c41b891dd1c67b71795cb6c456141
Previous check wouldn't make sense, cause the last offset in the data
could be that one that makes data balanced (and j is increased always
after iteration).
Change-Id: Ie9498d0ac9e3417d09b8b3043bf3281e7dfbf9db
Relying on any native implementation of window.KeyEvent is
unreliable. The two specs listed in the comments for this method
are not the same and Firefox now has DOM_VK_ENTER and
DOM_VK_RETURN which are equivalent to 13 and 14 respectively.
The key on the right side of my home row triggers 14 using
native window.KeyEvent in Firefox, where Chrome uses our
defined constants and returns 13. This fix ensures the same
values for all browsers.
Change-Id: I12b2f5d674bdf13526577cb81d0505b608f2846c
Don't suggest a "new page" (redlink) for values that are
strictly different but the same as a suggestion according
to mw.Title (e.g. input "foo" with an existing "Foo" page
should not suggest a redlink to inexistant page "foo").
Bug: 48476
Change-Id: I66f5fc56554984af58d6223dc6cd76b3ab9940bd
Fetch
* Added basic name expansion logic since bug 48663 is not yet
resolved in Parsoid.
* Fetch info from template data API.
Clean up (follows-up 97157a1c, de203cb8, 718db58)
* Remove redundant initialize method.
* Pass the entire target object to #getTemplateData
because `target.wt` is *not* the template name.
* Reorganise #createPage to be more logical and rename
to #createParamPage.
* Add todo comments for things currently missing.
* Implement error handling of promise to prevent UI from
going stale if the promise is not resolved.
* Fix documentation in ve.ui.MWCategoryInputWidget.
Change-Id: Ie0114a81eead86d7a3b3e3a7a5b10d25c457b524
If the commit message is empty we show a warning, and if the warning
is already visible we allow the user to proceed.
Bug: 47752
Change-Id: Idba707abaea8b08a94f7fa4d5bc5b1e35261a572
We need to listen to the same events as bytelimit. Specifically
we were missing keyup, which meant the counter could get stuck on
-1.
Bug: 47718
Change-Id: I3d4f3f8cd451bfb6acea19ee9baae7be60adcf15
Rather than using namespaced linmod attributes, store the preserved
HTML attributes in the .htmlAttributes property of the linear model
element, in a nested structure to allow for easier treatment of child
nodes. Also added attribute order preservation by storing attributes
as an object plus an array of keys.
ve.ce.Node.js:
* Remove html/* attribute synchronization. Doesn't make sense any more
because these things aren't in the attributes object any more. I don't
think it ever made sense because these attributes were never supposed
to be changed anyway.
ve.ce.View.js:
* Replace renderAttributes() with a simple wrapper around
renderHtmlAttributeList()
ve.dm.Converter.js:
* Add buildHtmlAttributeList() and renderHtmlAttributes() for building
and rendering HTML attribute lists
ve.dm.Model.js:
* Add getter for .htmlAttributes
ve.dm.Node.js:
* Drop .htmlAttributes on clone, and remove logic dropping html/*
ve.ui.MWCategoryWidget.js:
* Remove html/0/about hack, was already unnecessary and now doesn't
work any more
tests/:
* UPDATE ALL THE TESTS
Change-Id: I620573afd70d36ade6b80413075b6e1f4a435abe
Intercept badtoken errors, refetch the edit token from the
action=tokens API, and retry the request again. If this fails too,
show the error to the user.
Right now this just shows the good old confirm() dialog if the token
refetch fails; we should probaby give the user a clearer error message
telling them to refresh the page or something.
Bug: 42984
Change-Id: Ib43d1938ffa24bc8d1dc76a300e16e486dabd928
Add checks to make sure parts of the target are acutally
active before trying to tear them down or use them.
This fixes a couple of issues which appeared when the VE was
closed before it had a chance to finish loading.
Bug: 48520
Bug: 47813
Change-Id: Ide5def0e983bab49108b40008fc170957c7fc2a2
Objective:
* Context popup would stop opening sometimes "mysteriously" which ended
up having to do with the automatic closing on blur functionality
added to popups for use in the category popup widget
* Mousedown event canceling was being applied a little too widely, and
was causing popup widgets to not allow child elements to be focused
(unless they contained an iframe, like an inspector)
Changes:
ve.ui.Context.js
* Make use of the popup's show and hide methods within the inspector
ve.ui.MWCategoryPopupWidget.js
* Override autoClose option for category popup widgets
ve.ui.PopupWidget.js
* Add autoClose option to popup widgets
* Move event handler to the top of the methods (convention)
* Only bind blur event if autoClose is enabled
* Inline the getFocusedChild method
Change-Id: I22aedb5fbd51b327ea7ce2ecdd6123e79cbebb9c
Objectives:
* Make the context menu display in the top right corner of the currently
focused inspectable node (if there is one)
* Prevent clicking on anything to do with the toolbar or popup from doing
anything at all, ever
Bonus:
* While we are using the clever feature in jQuery's on method which allows
passing boolean false to cancel the event - may as well do that in
ve.ui.Dialog as well
Changes:
ve.ui.FocusableNode
* Add ability to specify the focusable element so that dimensions can be
derived from it
ve.ce.Surface
* Add quotes to object keys
ve.ui.MediaDialog
* Change association from being MW specific to handling images in general
ve.ui.Context
* Add embedded styles for context
* Add embedded mode, which is triggered when the context is a single
focusable node, and the node is large enough to fit the context
reasonably
ve.ui.Dialog
* Inline mousedown handler
ve.ui.Toolbar, ve.ui.PopupWidget
* Cancel stray mousedown events
Change-Id: I4b25d33f64b4bcb8a3ecfd7e9728f54a2d4886f3
When loading an image block node in the demo, there are no classes on the figure, which caused the code to crash because it was assuming that jQuery's attr method would always return a string.
Change-Id: Ib13e7bfa3fb2bd76ac71dfef085bed209d880b4a
