fix(timepicker)!: errors following a full review of the component (#DS-5482)

[artembelik]

What

A full review of timepicker, in the same shape as the 20.3.0 component reviews.

What stays a plain accessor, and why

KbqTimepicker implements KbqFormFieldControl, which declares value, id, placeholder, required, disabled, focused, empty and errorState as plain members. Those stay accessors — the interface is how the form field reads them. The six of them carried // TODO: Skipped for migration because: comments from the automated tooling; they say the actual reason now.

min and max reported something other than what was bound

set min(value: D | null) {
    this._min = this.getValidDateOrNull(this.dateAdapter.deserialize(value));
    this.validatorOnChange();
}

A value the date adapter could not read came back as null from the getter. They report what was bound now; the parsed values are internal computeds that still drive the validators, re-run from an effect.

kbqValidationTooltip leaked a subscription per binding

set kbqValidationTooltip(tooltip: KbqTooltipTrigger) {
    …
    this.incorrectInput.subscribe(() => { … });   // never unsubscribed
}

Re-binding the input stacked another subscription, and the last one outlived the directive. It is a signal input driven by an effect with an onCleanup teardown.

format's side effects were already covered

Its setter clamped the value, refreshed the placeholder and re-rendered the view. The clamp is the input's transform; the other two are subsumed by the locale effect that already re-read timeFormatPlaceholder — which reads format(), so it re-runs on a format change too.

That effect also stopped returning early on a consumer-provided placeholder. It used to skip the value reformat along with the placeholder update, so a locale change left the rendered time in the old locale whenever the consumer had set their own placeholder.

The existing "Display time corresponding to timeformat" tests — including the unsupported-format fallback and the format-change re-render — pass unchanged, which is the evidence that moving the side effects preserved behavior.

Also

• The generated id comes from the CDK _IdGenerator instead of a module counter (kbq-timepicker-1 → kbq-timepicker-a1), and this.id = this.id in the constructor is replaced by initialising the backing field.
• disabled and required declare booleanAttribute instead of calling coerceBooleanProperty inside their setters.

Migration

timepicker-signals runs from ng update @koobiq/components@20. It rewrites format reads and reports the rest. min and max are deliberately not rewritten: appending () would compile and hand back a different value for anything the adapter could not parse.

There is no template pass — kbqTimepicker is an attribute on a native <input>.

Documented in docs/guides/migration.{en,ru}.md, section 18.

Testing

• timepicker.spec.ts: 64 → 67 tests. New coverage for the format clamp through the transform, min/max reporting the bound value, and the validators re-running when min changes.
• timepicker-signals/index.spec.ts: 10 tests.
• Full packages/components (5164 tests) and packages/schematics (492 tests) suites pass.
• check-api is in sync.

BREAKING CHANGE

🤖 Generated with Claude Code

[github-actions]

Visit the preview URL for this PR (updated for commit 8ba92c0):

https://koobiq-next--prs-1992-as4benoa.web.app

_{(expires Mon, 07 Sep 2026 07:37:51 GMT)}

_{🔥 via Firebase Hosting GitHub Action 🌎}

_{Sign: c9e37e518febda70d0317d07e8ceb35ac43c534c}