angular / components

Removes a remaining instance of the Bazel `module_name`
property being set.

The API goldens have been flattened as part of 766b07e,
but the script for updating goldens has not been updated.

This commit fixes the script so that it respects the new file location
of goldens.

We recently introduced a new tool for testing the public API
signature of entry-points. This tool is part of the shared
dev-infra package and uses Microsoft's API extractor under the
hood. This commit sets up the new tool and removes ts-api-guardian.

Updates the Bazel NodeJS rules to v4.0.0-beta.0. This is necessary
so that the Angular components repo can start using the new API
golden tool from the shared dev-infra repository, and it's generally
good to stay as up-to-date as possible with the Bazel rules as it's
easy to fall behind, and updating early allows us to discover issues
affecting our tooling earlier (where they are easier to address due to
e.g. potential breaking change policy).

bazelbuild/rules_nodejs@3be2902.

As part of this change, we remove all manual `module_name`
properties from `ts_library` and `ng_module` targets. These
will be generated now as manually setting the module name is
rather prone to mistakes and this way it also integrates better
with the new `package_name` property introduced as part of
Rules NodeJS v4. This change reveiled that we set a module name
for the private `material/testing` folder. This folder no longer
will resolve through `@angular/material` and needs to be imported
through a relative path.

* fix(material-experimental/mdc-table): match legacy examples

* fix(material-experimental/mdc-table): fix mdc css class

Fixes a bug in the example of Angular Material `mat-table` with columns generated via ngFor.
The fix changes the tags over to native table selectors, which brings the example into alignment
with other examples of `mat-table` usage and allows for Angular Material formatting to apply.

#23169)

Fixes an error that is thrown, because the MDC mixin we were using was renamed.

#23147)

We have some logic in `mat-autocomplete` that scrolls the list to the top when the user moves to the first option in the first option group. This is slightly better UX, because it shows the group label, rather than stopping just below it.

These changes port over the same logic to `mat-select`.

…23106)

* fix(material-experimental/mdc-tooltip): make background color opaque

* due to a11y concerns surrounding the 0.6 background color opacity of the
  mdc-tooltip, we are switching to using our own custom background color
  instead of the one defined in the MDC styles.

* fixup! fix(material-experimental/mdc-tooltip): make background color opaque

* fixup! fix(material-experimental/mdc-tooltip): make background color opaque

…pace (#23136)

In #19700 a flag was introduced whose purpose was to indicate whether focus should be moved from the input to the chips when pressing backspace. The problem is that the flag was only being updated on key presses and on init which means that the state will be incorrect if the value changes programmatically after init.

Fixes #23128.

…#23126)

Fixes that the CDK drag&drop could be triggered by fake `mousedown` or `touchstart` events dispatched by a screen reader.

#22066)" (#23152)

This reverts commit ae968ab.

…lex tables (#22350)" (#23151)

This reverts commit c76a09e.

#23150)

This reverts commit 71d75f4.

We transition the circle of a radio button to `scale(0.001)` in order to hide it and to work
around an animation in IE. It seems that on higher system zoom levels (e.g. 125%+) the
browser approximates the size to 1x1 which can be visible.

These changes work around the issue by also setting `opacity: 0` while the circle is inactive
and isn't animating.

Fixes #22036.

* fix(cdk/table): set default role of table to 'table'

* fixup! fix(cdk/table): set default role of table to 'table'

…es (#22350)

In an earlier PR we switched the flex-based table styles to target the tag names, rather than classes in order to support flex tables in MDC. The problem with targeting tag names is that it doesn't allow our styles to be applied to other elements within the table. E.g. in #22349 the user isn't able to target the "no data" row.

These changes switch back to targeting classes and add some styles to prevent the flex-specific styles from bleeding into the table-based ones.

Fixes #22349.

Fixes that the list animations weren't being disabled with the `NoopAnimationsModule`.

…23102)

Under the Accessibility section of the mat-paginator it's written that
a `<mat-paginator>` must have an aria-label.
But actually none of the found examples does provide an aria-label.

https://github.com/angular/components/blob/master/src/material/paginator/paginator.md#accessibility

Most of the exports we've excluded from the exports config are underscored APIs. These changes add an option that allows us to exclude exports based on on a regex and removes the underscored APIs from the config since they're private anyway.

Makes the schematics code compliant with the `strictFunctionTypes` compiler option.

There was a typo in the `cdkDragPreviewContainer` input name in the docs.

Fixes #23114.

…#23125)

Adds support for two-way data binding on the `selected` property of `mat-list-option`.

Fixes #23122.

We don't need to re-measure the viewport any time it is resized, we just need to clear the cache so that it is re-measured next time somebody requests it.

…adow dom (#23121)

We were repeating the logic for resolving the event target while accounting for shadow DOM in a few places. These changes add a common utility instead.

Fixes that we didn't have any styles that would differentiate a disabled item from an enabled one.

Similarly to other input-based components, these changes add an `aria-describedby` input to the slide toggle.

Fixes #23094.

…23095)

Fixes that the `color` type inputs didn't work due to a style override we had on top of MDC.

Aligns the ripple timings with the numbers from MDC.

…23040)

Fixes that the MDC-based slider didn't have a strong focus indicator.

The Material table with sticky headers sets its `position` as `!important` which ends up overriding the dragging styles. These changes rework the internals to use `setProperty` instead of `element.style.position` which allows us to set the `position` as `!important` as well. It also has the advantage that we can write the properties as dash case, instead of having to guess whether the vendor-prefixed properties are camel case or pascal case.

Fixes #22781.