* Creates a CDK schematic that can be used to generate a component that uses the CDK Drag & Drop module. The code is based on the `sorting` example of the drag-drop (just a bit trimmed to keep the code simple)
* Updates the Bazel TypeScript rules.
* Fixes a test case which doesn't match the name of the upgrade data it runs with (`css-names` --> `css-selectors`)
* Removes Bazel version check (the NodeJS rules automatically check the Bazel version based on the version of the NodeJS rules)
* Currently constants which are exported, will not show up in the docs. This introduces support for displaying `constants` in the docs.
* Also creates a Nunjucks plugin that supports highlighting specific code snippets. This is helpful when rendering a type in the docs (similar to angular/aio)
* Marks most animation constants as `@docs-private` because those shouldn't be relevant to users.
* Currently if someone toggles the slide-toggle through mouse interaction, the persistent ripple will show up besides one transient ripple. The persistent ripple should only show up if the native checkbox element was focused through keyboard.
Fixes#13291
Allows for consumers to project extra content into the label of `mat-optgroup`. Currently any content that isn't a `mat-option` won't be rendered.
Fixes#11489.
Looks like something that slipped through when refactoring the previous PR to support multiple appearances. The selector that was supposed to set the proper borders in vertical mode was incorrect.
Currently we only listen for keydown events on the autocomplete trigger, however because the escape handler doesn't go through the `OverlayKeyboardDispatcher`, it means parent overlay will still receive the event, causing unintended side effects like the parent dialog closing when escape is pressed on an autocomplete. These changes switch the closing actions to go through the `OverlayRef.keydownEvents()`.
* Adds a drawing which has been created before implementing the shared schematic logic with the CDK. This drawing should be in the repository because it avoids recurring questions about the CLI workflow.
* Due to the upgrade to TypeScript 3.0.0 and the fact that we refer to the actual sources instead of `../../../dist/packages/cdk` (for path mappings), TypeScript will output the Material schematics in an incorrect output structure. This can be fixed by taking advantage of the new "Project References" feature.
https://github.com/angular/material2-builds/tree/master/schematics/lib/schematics (Notice the additional `lib/schematics` folder in the `@angular/material/schematics` folder).
Along the same lines as #13361. Fixes the selection list not marking all selected options correctly when an array with duplicate values is assigned programmatically.
Doesn't type the `SelectionModel.onChange` event to `Subject | null` anymore, in order to avoid having people null-check it on every consumption.
Fixes#12347.
Fixes the `selectionChange` event always firing for the setter, `select`, `deselect` and `selectViaInteraction`, even if the value hasn't changed. Also gets rid of some repetition around emitting the event.
The offset of a sticky row is calculated by summing up the heights of
the previous sticky rows. As the access to the height causes a forced
reflow this is now skipped when there is only one row.
Fixes the `resize` and `orientationchange` events from the `ViewportRuler` being run inside the `NgZone`, even though the intention is that they shouldn't be. The issue comes from the fact that we call `merge` outside the `NgZone`, however the thing that causes the event to be bound is the subscription right after it.
Fixes#12883.
Currently when a focus trap is disabled, we set the `tabindex` of the anchors to -1 in order to allow people to tab out of it. This doesn't work if somebody is navigating with the arrow keys using a screen reader, because the element is still focusable which means that the screen reader will focus it eventually, causing focus to be trapped. These changes remove the `tabindex` if the focus trap is disabled instead.
**Note:** An alternate approach to this can be to hide the element using `display: none`, but I opted to remove the `tabindex` in order to avoid a style recalculation.
Fixes#13132.
Currently `mat-menu` displays focus indication only if it has been opened by keyboard. We detect how the menu was opened using a `mousedown` event. The issue comes from the fact that `mousedown` also fires for right or middle button clicks, event though a subsequent `click` won't fire, which will throw off our detection if the user opens the menu with their keyboard afterwards. These changes add a check that ensures that only left button clicks are considered as opening.
Fixes a centered flexible overlay with pushing, on a scrolled page, not calculating the position properly. There were a couple of issues:
* We were using the `top` viewport offset to calculate along the X axis, as well as `left` to calculate along Y.
* We weren't accounting correctly for the scroll position.
Fixes#11868.
Resolves a long-standing TODO about starting the drag&drop sequence after the user has dragged over a certain amount of pixels, rather than starting it immediately on `mousedown`/`touchstart`. This prevents random clicks and taps on the element from being interpreted as dragging and avoids potentially interrupting the native click events on children of `CdkDrag`.
Currently we remove the `panelClass` as soon as `detach` is called, however the pane may still be animating. These changes wait for the panel to be done animating and to be removed from the DOM, before removing the class.
Fixes#13189.
With TypeScript 3.1, several native DOM types have been updated. Notably, `document.documentElement` and `document.activeElement` are now nullable. APIs relating to `matchMedia` were also updated.
Adds the `matBadgeDisabled` input that allows for a badge to be styled as if it was disabled. This allows it to blend in when it's attached to an element that is also disabled.
Fixes#13191.
Since we count on the options being a particular height, having 1px dividers between them ends up throwing off the keyboard navigation. These changes add an offset to the dividers to avoid the issue.
Fixes#13200.
Currently we set `aria-invalid` automatically on inputs inside a form field, but not on one inside a chip list. These changes add `aria-invalid` to the chip input, when the related chip list is invalid.
Fixes#13205.
Along the same lines as #13308. Currently we forward the `tabindex` from the host to the underlying `input`, however we never clear the `tabindex` from the host, which means that we end up with two layers of focusable elements.
* Since the `webdriver-manager` always tries to query the versions of all Selenium drivers, we currently run into rate limit failures that cause the E2E tests to not run properly.
Joey Perrott