fix(textarea)!: errors following a full review of the component (#DS-5482) - #1990

What

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

What stays a plain accessor, and why

KbqTextarea implements KbqFormFieldControl, which declares value, id, placeholder, required, disabled, focused, empty and errorState as plain members. That interface is how the form field reads them, and it went through its own review in 20.3.0 — so those stay accessors. What moved are the four inputs the textarea owns.

canGrow disagreed with what was bound

get canGrow(): boolean {
    return !this.maxRowLimitReached && this._canGrow;
}

It reported false once the textarea hit maxRows, even though the consumer had asked for growth. The folded value is what drives the resize handle, so it became an internal growing computed; canGrow() reports what was bound. Same shape as KbqLink.tabIndex in the earlier review.

freeRowsHeight defaulted itself by writing its own input

this.freeRowsHeight = this.freeRowsHeight ?? this.lineHeight;

That is exactly why the automated migration skipped it. The fallback is a computed over the measured line height now, so binding [freeRowsHeight] later actually takes effect instead of being overwritten the next time ngOnInit runs — and it does run again, because the parent animation re-invokes it.

maxRows and freeRowsHeight lied about being required

Both were declared number while an unbound textarea held undefined. maxRowLimitReached compared rowsCount > undefined, which is false — that accident is what made unlimited growth work. They report number | undefined now and the comparison is explicit.

The row-limit class waited for someone else's change detection

rowsCount is written inside runOutsideAngular, so [class.kbq-textarea_max-row-limit-reached] only appeared on the next unrelated change detection pass. It is a signal now and the class follows it directly.

Also

• The parent.animationDone subscription had no teardown; it is takeUntilDestroyed() now.
• The generated id comes from the CDK _IdGenerator instead of a module counter, so the shape changes from kbq-textarea-1 to kbq-textarea-a1.
• this.id = this.id in the constructor — a trick to force the setter — is replaced by initialising the backing field.

Migration

textarea-signals runs from ng update @koobiq/components@20. It rewrites the value-safe reads and reports the rest. canGrow is deliberately not rewritten: appending () would compile and hand back a different boolean at the row limit.

There is no template pass — kbqTextarea is an attribute on a native <textarea>, so a reference variable is not tied to an element name the schematic can match.

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

Testing

• textarea.component.spec.ts: 24 → 27 tests. New coverage for the valueless canGrow attribute, canGrow() reporting the bound value, and the unbound maxRows/freeRowsHeight defaults.
• textarea-signals/index.spec.ts: 10 tests.
• Full packages/components (5164 tests) and packages/schematics (492 tests) suites pass.
• check-api is in sync.

The row-limit reactivity fix is not unit-testable here: jsdom reports no line height, so rowsCount never leaves 0.

BREAKING CHANGE

🤖 Generated with Claude Code