When using two-way binding with isolate scope, under some circumstances
the lastValue variable captured in the parentValueWatch function can get
out of sync.
Specifically, if both the value in the origin scope as well as the value
in the isolate scope get independently updated to the same value within
one digest cycle, the lastValue is never updated. This potentially causes
the watch to make the wrong decision as to which side to update on subsequent
passes.
This fixes things by ensuring lastValue is always set to the last seen
value even if the watch's logic was short circuited because there was no
difference between the values in the original and isolate scopes.
Closes#5182
innerText depends on styling as it doesn't display hidden elements.
Therefore, it's better to use textContent not to cause unnecessary
reflows. However, IE<9 don't support textContent so the innerText
fallback is necessary.
This version matches the "alternate 2.2" version here: http://jsperf.com/ensuresafeobject/2
alternate 2.3 is a bit faster and simpler, but would break backwards compatibility.
Closes#5246
When a jqLite collection contains text nodes, find() does not work :-(
This fix ignores all nodes than can't do getElementsByTagName()
It seems a little bit faster than testing nodeType : http://jsperf.com/nodetype-vs-duck-typingCloses#4120
In Safari 7 (and other browsers potentially using the latest YARR JIT library)
regular expressions are not always executed immediately that they are called.
The regex is only evaluated (lazily) when you first access properties on the `matches`
result object returned from the regex call.
In the case of `decodeEntities()`, we were updating this returned object, `parts[0] = ''`,
before accessing it, `if (parts[2])', and so our change was overwritten by the result
of executing the regex.
The solution here is not to modify the match result object at all. We only need to make use
of the three match results directly in code.
Developers should be aware, in the future, when using regex, to read from the result object
before making modifications to it.
There is no additional test committed here, because when run against Safari 7, this
bug caused numerous specs to fail, which are all fixed by this commit.
Closes#5193Closes#5192
The urlResolve method was fixed to automatically remove the
volume label from path names to fix issues with the file
protocol on windows where $location.path() was returning
paths where the first segment would be the volume name,
such as "/C:/mypath". See #4942 and #4928
However, the solution was specific to the $location non-
HTML5 mode, and was implemented at a lower level of
abstraction than it should have been. This refactor moves
the fix to inside of the LocationHashBangUrl $$parse method.
Closes#5041
`$sanitize` now uses the same mechanism as `$compile` to validate uris.
By this, the validation in `$sanitize` is more general and can be
configured in the same way as the one in `$compile`.
Changes
- Creates the new private service `$$sanitizeUri`.
- Moves related specs from `compileSpec.js` into `sanitizeUriSpec.js`.
- Refactors the `linky` filter to be less dependent on `$sanitize`
internal functions.
Fixes#3748.
Previously if an app was running from file:// origin we would always return either
http 200 or 404 depending on whether the response was present.
This changes the behavior so that we do this only if the protocol of the request
(not the origin) is file:// and only if the status code is 0.
Closes#4436Closes#4587Closes#4514
Although modern browser support the "input" event, they still only fire
the "change" event when they auto complete form elements
other than the currently selected one.
Related to #1460
IE8, IE9 and IE10 can use `script.onreadystate` so up till now we have been using this
if the sniffer says we are on IE.
But IE11 now does not support `script.onreadystate` and only supports the more standard
`script.onload` and `script.onerror`.
IE9 and IE10 do support `script.onload` and `script.onerror`. So now we only test whether
we are on IE8 or earlier before using `script.onreadystate`.
See http://pieisgood.org/test/script-link-events/
jQuery just uses all these handlers at once and hopes for the best, but since IE9 and IE10
support both sets of handlers, this could cause the handlers to be run more than once.
jQuery also notes that there is a potential memory leak in IE unless we remove the handlers
from the script object once they are run. So we are doing this too, now.
Closes#4523Closes#4527Closes#4922
Previously, when unwrapping promises was set to `true`,
an error would occur if a parsed expression had a
new line in it.
This was because when generating the `evaledFnGetter` code,
a new line in an parsed expression would create a new line
in a JS string in that code, which is illegal. That is:
```js
pw("A+
B")
```
Closes#4718
When a request is aborted, it makes no sense to read the response headers or text.
Also in IE9, trying to read data (either response headers or text) from an aborted request
throws an Error c00c023f.
Fixes#4913Closes#4940
Fixes an issue with httpBackend expectations where a given body object
may not match the actual request body if its keys are serialized in a
different order.
Closes#4956
This is highlighted in angular-phonecat when you try to use the index-async.html
which needs to load the ngResource module asynchronously but fails when it tries
to call `angular.$$minErr` to create the $resourceMinErr object.
Closes#5050
Keyframe animations trigger on the first CSS class and not the second.
This may cause a slight flicker during a stagger animation since the
animation has already started before the stagger delay is considered.
This fix ensures that the animation is blocked until the active animation
starts which allows for staggering animations to take over properly.
Closes#5018
Transitions are blocked when the base CSS class is added at the start of the animation. This
causes an issue if the followup CSS class contains animatable-styles. Now, once the animation
active state is triggered (when the animation CSS dom operation occurs) the animation itself
will always trigger an animate without a quick jump.
Closes#5014Closes#4265
When $compile interpolates a CSS class attribute expression it will
do so by comparing the CSS class value already present on the element.
This may lead to unexpected results when dealing with ngClass values being
added and removed therefore it is best that both compile and ngClass delegate
addClass/removeClass operations to the same block of code.
Previously, calling `MyResource.save(myResourceInstance)`returned
a promise, in contrast to the docs for `$resource`. However,
calling `MyResource.save({name: 'Tobias"})`already correctly
returned a resource instance.
Fixes#4545.
Closes#5061.
The clip property seems to remove the box-shadow property when an absolute
positioned animation is ongoing. This fix changes the property to be border-spacing
which is also very underused. The border-spacing CSS property is only visible
when border-collapse is set to separate.
Closes#4902Closes#5030
ngClass works by removing all the former classes and then adding all the
new classes to the element during each watch change operation. This may
cause transition animations to never render. The ngClass directive will
now only add and remove the classes that change during each watch operation.
Closes#4960Closes#4944
Depending on the animations placed on ngClass, the DOM operation may
run twice causing a race condition between addClass and removeClass.
Depending on what classes are removed and added via $compile this may
cause all CSS classes to be removed accidentally from the element
being animated.
Closes#4949
In 1.2, the behavior of ngInclude was modified to use DOM APIs rather than jqLite. This means that
even when jQuery was loaded, ngInclude was not calling into it, and thus scripts were not eval'd
as they had been before. Although the use of ngInclude to eval scripts as a lazy-loading strategy
was never an intentional feature, this patch restores the ability to do so.
Closes#3756
