Textarea

Contentless host-mediated multiline text-control prototype

base-textarea-root is a contentless Base prototype. It statically declares a plain-text, multiline, host-owned editing requirement and lets the adapter select one physical host editor. The current Web profile resolves that requirement to a native <textarea>; the Web Component, React, and Vue demos are cross-adapter evidence on one Web host rather than completed multi-host conformance.

Value ownership is selected when the instance is created and remains stable for its lifetime:

• With defaultValue, Root is uncontrolled. It seeds its value once, adopts subsequent user input, and keeps dirty input when unrelated props update.
• With value, Root is controlled. User edits are proposals: Root emits them, but the owner must pass an accepted value back before it becomes the authoritative value.

The outward events contain normalized data rather than native event objects:

• valueChange follows native input and emits { value, composing, data, inputType }.
• change follows the native change boundary and emits { value }.
• compositionStart, compositionUpdate, and compositionEnd each emit { value, data } for IME observation.

During a controlled IME session, the physical textarea keeps the in-progress candidate visible. After compositionEnd, an unaccepted candidate is restored to the current controlled value; restoration never interrupts composition.

Ordinary property updates and synchronization with an unchanged value preserve the current cursor and selection. Long-press selection, selection handles, software keyboards, and system copy, cut, and paste menus belong to the host editing session. Base Textarea neither defines a touch-only prototype nor replaces system editing chrome.

Root projects these props to the same physical textarea: disabled, readOnly, required, name, placeholder, autoComplete, minLength, maxLength, rows, and wrap. value and defaultValue participate in the ownership rules above rather than creating a second control.

The expose surface includes focusSelf() and blurSelf() plus value, disabled, readOnly, focused, focusVisible, and composing state. Accessibility projection keeps the textarea’s textbox role, mirrors disabled/read-only state to aria-disabled/aria-readonly, and maps ariaLabel, labelledBy, and describedBy to aria-label, aria-labelledby, and aria-describedby on that physical target.

Base contributes no child renderer and no visual treatment. App authors style and label the native control around this contentless protocol boundary rather than placing authored or rich-text content inside Root.

asTextareaRoot.modules publishes the authored asHook’s frozen pre-render module requirements. A styled prototype using asTextareaRoot() must reuse this surface on its prototype definition; the setup-time call cannot retroactively change adapter target selection.

base-textarea-root does not own form submission, hidden fields, form-data coordination, validation messaging, auto-resize, rich-text editing, status/live-region announcements, system selection or edit-menu chrome, or composition outside the single physical editor. Web form-related props can still be projected without making the prototype the owner of a form workflow.

• Prototype: P-BASE-TEXTAREA
• Contract: C-TEXT-CONTROL-0001
• Source: packages/prototypes/base/src/textarea
• Test: textarea.test.ts