This fixes the results display so they prefer the `selected`
property on objects over checking if the `id` matches the value of
the underlying select.
We only ignore the `selected` property if it is null, at that point
we compare against the `id` property.
This closes https://github.com/select2/select2/issues/3163.
This also fixes an issue where we were using the `full_name` for
the value instead of the `id` in the example, which technically
wasn't correct.
This closes https://github.com/select2/select2/issues/3148.
In the past, `initSelection` was used for loading in data objects
for the pre-selected options. Now that Select2 is using a `<select>`
element, there is no need for doing this because the `<option>`
elements should provide the relevant information. So you can now
just pull this information from the remote data source
(or static array, in some cases) and build out the `<option>`
elements manually.
In most cases you don't need to go the full length with a custom
data adapter, but instead are just looking to pre-load elements on
the initial page load.
This improves https://github.com/select2/select2/issues/3116.
Now we can use relative paths in the calls to `require`, so those
using relative builds don't have to worry about modules not being
imported properly. This also makes it considerably easier in the
future when we look to cut down the build sizes and have a custom
builder, as we don't have to worry as much about the paths being
incorrect.
This fixes https://github.com/select2/select2/issues/3014 finally.
Now we do not have separate AMD builds, instead we use a UMD wrapper
to support both AMD and CommonJS environments along with the browser.
This has been tested with RequireJS and it correctly imports Select2
and hooks into jQuery.
This means that all builds will include the Almond AMD loader, but
this isn't that much of an issue when you consider the file sizes.
The loader is namespaced so RequireJS and other AMD builders don't
complain about `require` and `define` existing. This is also done
so AMD loaders don't conflict, as Select2's AMD loader should
operate independently of others.
There are now two layers of wrappers that are applied to files,
only one of which is used for the translation files. This allows
us to only load in the internal Select2 modules once, and in
general makes most things cleaner. These have also been pulled out
into individual JS files instead of being stored in JSON files,
allowing us to add comments and make things look readable.
Because we use a few hacks, these files are heavily commented as well.
This closes https://github.com/select2/select2/issues/2945.
This closes https://github.com/select2/select2/issues/2946.
This finally fixes and adds some integration tests to make sure
that `select2('data')` works as we are expecting. This also adds
tests for `select2('val')` and fixes undefined variable issues
that were spotted because of these tests.
We also no longer date stamp distribution builds.
This closes https://github.com/select2/select2/issues/3104.
This moves the clear icon to the top of the DOM from the bottom to
ensure that it is always visible on top of the selection.
This closes https://github.com/select2/select2/issues/3089.
We don't need to update the dates on these every time that grunt
is called. This gives it a much more basic banner that is similar
to the minified version.
The nested data attributes are only supported on jQuery 2.x or
browsers which support the `dataset` attributes on DOM elements.
In order to prevent test failures, and because tests cannot yet
be skipped conditionally, we just cut the test early.
As jQuery 1.7.2 is the lowest version of jQuery supported by Select2,
it makes sense to run the tests on it. For the most part, we can
assume that the newer versions of Select2 are backwards compatible
enough such that this isn't an issue.
The recommended version of jQuery to use is the latest though, which
is why the jQuery file is only included in the tests.
This revealed a few issues with our data fallbacks and `.append`
functionality that was introduced in jQuery 1.8.
Looks like Sauce Labs is not going to have tests running on later
versions of Opera, probably because it switched to Chromium/Blink.
Removing it from the builds so we reduce the time it takes.
This strips whitespace in tags by default, so multiple tags cannot
be created with only whitespace as the difference in the id.
A test has been added to ensure that this remains fixed in the future.
This closes https://github.com/select2/select2/issues/3076.
This fixes an issue with `select2('data')` that caused it to not
work as intended. Because we were using `Select2.data` for the
data adapter, it was overriding the `Select2.data()` method which
was supposed to be used.
This just renames the `Select2.data` property to
`Select2.dataAdapter` to prevent the naming collision.
This closes https://github.com/select2/select2/issues/3104.
This fixes an issue when using a `<select>` where the elements were
created with XHTML-encoded characters to prevent any injection, as
they would be double-encoded and display incorrectly.
When using a `<select>`, we can assume that the data has already
been encoded because any XSS will have already run before we get to
it. Because of this, we can just use `.text()` instead of `.html()`
to avoid any issues.
This also includes a test to ensure that this does not become an
issue in the future.
This closes https://github.com/select2/select2/issues/3115.
This fixes an issue where the dropdown would not automatically
open when something was typed in the search box on a multiple
select. This was most noticeable when the dropdown closed and
the search box was focused and the user started to type, as the
text would appear in the search box, queries would be sent out to
retrieve results, but the dropdown would not be opened to display
the results.
This also fixes an issue introduced in a recent commit where search
queries would be sent out twice on modern browsers that supported
both the `keyup` and `input` event. The `keyup` event is now
properly debounced for these browsers and the queries are only
going out once.
This closes https://github.com/select2/select2/issues/3036.
Previously, when in results the enter key would select items that
were highlighted if they were not already selected. In the case of
a select where multiple items could be selected, pressing enter
when highlighting a selected item would also allow it to be
unselected. While this seems intuitive for accessibility purposes,
the enter button essentially working as a toggle, it caused some
really strange behavior.
- If the enter button was held down, all previously selected items
would be unselected.
- The enter button did not work the same across both single and
multiple selects.
After listening to user feedback, I have decided to remove the
"enter as toggle" functionality from Select2 and have gone back to
just having the enter button select items. This means that instead
of unselected items that are already selected and highlighted,
Select2 will just close the dropdown. This is the same as what
Select2 would previously do for single selects, so the keyboard
functionality is now the same across both.
Because this removed the only easy way to unselect items in the
dropdown using the keyboard, we had to maintain the toggle
functionality. We decided to implement the toggle functionality
on the CTRL + Space keybinding, which is in line with other
applications. Now when pressing CTRL + Space at the same time in
the dropdown, the highlighted result will behave the same as if the
mouse selected it, which will toggle the current item in multiple
select mode and close the dropdown in single select mode.
This is the same keybinding that Windows Explorer [1] and GTK [2]
use for toggling the current selection, which was why it was picked.
This also fixes an issue where keyboard focus would be lost once an
item was unselected from the results. This was due to a bug in the
CloseOnSelect module that would only automatically close the
dropdown when an item was selected, but not when an item was
unselected. Now the dropdown will be closed automatically when an
item is unselected, which will also cause the selection (and
eventually the search) to be focused.
This fixes two issues described in
https://github.com/select2/select2/issues/3036#issuecomment-76321411.
[1]: http://superuser.com/q/78891/72528
[2]: https://developer.gnome.org/gtk3/stable/GtkIconView.html#GtkIconView-toggle-cursor-item
This fixes an issue where most keys would not trigger the search
in Firefox for Android. There were only a few keys which would
trigger `keyup`, such as space and enter (the search icon), but
they were not consistent. We know that they were being triggered
though, as you could type "new " (note the space) and it would
trigger a search, giving us all states that started with "new".
The problem is that Firefox for Android does not consistently
trigger `keyup` and `keydown` events when a keyboard is used. To
work around the issue, we are now using the `input` event in
replacement of the `keyup` event, which was used to trigger the
search. While this is not an actual `KeyboardEvent` in Chrome, and
lacks some of the important metadata such as `which`/`key` in
Firefox, it works for our implementation.
As the `input` event is not supported in older browsers, such as
those before Internet Explorer 9, we have to listen for both the
old `keyup` event and the new `input` event. As the `input` event
is always triggered before the `keyup` event, we unbind the `keyup`
event automatically to prevent searches from being triggered twice.
This solution was discovered in a blog post by Mathias Bynens at
https://mathiasbynens.be/notes/oninput.
**Note:** The ability to backspace in a blank text field on multiple
selects in order to remove the last selected item does not work in
Firefox for Android because the `keydown` event does not trigger
when the text field is empty. Users can still use the "x" icon
provided at the start of every selected option to achieve the same
effect.
This closes https://github.com/select2/select2/issues/2997.
This also fixes a possible issue where the `< 0` fallback would
not be used because the `minimumResultsForSearch` check would
always return false and never load in the module.
This closes https://github.com/select2/select2/pull/3077.
This fixes the fallback path for the `data-ajax-url` attribute on
elements. As this attribute was previously supported in Select2,
the attribute has been migrated to the new, nested format of the
url and triggers a deprecation warning when it is used. Because
of a fix to the `data-*` attribute parsing made in a9f6d64 that
allowed for nested attributes to be parsed correctly in modern
browsers under jQuery 1.x, the deprecation warning would be
triggered but the attribute would no longer actually be used.
This also fixes some of the `.data` calls to use the camel cased
version of the key instead of the dashed version, which is the
preferred key and will be enforced in future versions of jQuery
as the only way to access data attributes.
Now in situations where the `dataset` attribute is used by Select2,
it combines the results of both `$e.data()` and `e.dataset` when
generating the object containing all of the options. This will
the `dataset` fix to still be used, while also still relying on
jQuery to do additional parsing on any options that it can.
The `dataset` fix is now only used on jQuery 1.x, as that is the
only version of jQuery affected by the dash issue. This is done
using version number parsing on the `$.fn.jquery` property that is
defined by jQuery. As this property is not defined in Zepto and
many other jQuery compatible checks, we only include the fallback
if the property is available. This assumes that any jQuery
compatible libraries that are in use will not include the same dash
issue, which we believe is a safe assumption given that it did not
match the HTML `dataset` specification.
This also adds a few tests to ensure that the deprecated attributes
still continue to function.
This closes https://github.com/select2/select2/issues/3086.
