canvas-pilot/docs/development/design/mapping-window.md

5.0 KiB
Raw Blame History

Mapping window — visual study 01

Open the standalone mockup · Development

Status: proposal for visual review, not an implemented application window. Open the HTML file locally in a browser. It contains sample data, inline CSS and native controls only: no JavaScript, external assets, storage, network requests or device commands. Dropdown choices do not update other fields or counts.

The idea

Choose an editor layer on the left. Read its routes from top to bottom: one device source → this editor layer → several device destinations. The source means the device layer from which geometry/appearance is loaded, not its video input or signal routing. The actual Load and Send controls stay outside this mapping editor; Apply mapping would only save routing choices.

The first connection is the proposed initial connection for a new load mapping; that does not imply automatically choosing a device layer or performing a load. A load source has three explicit parts: connection, device screen and device layer. It can be absent independently of the send targets.

Send targets are grouped by connection, with screen/layer rows underneath. Repeat an editor layer across different layers on one screen or across screens without repeating the editor controls. Multiple connections are shown only as a collapsed future example; their implementation is not part of this study.

The useful shortcut is Follow load source on the first send target. This is a proposal, not a new saved setting: it would preserve today's simple synchronized workflow while allowing extra independent destinations. Unlinking it would keep its last destination, then allow editing it separately. Whether this should be a persistent relationship or an explicit “Copy load source” action needs agreement before changing the schema. Removing a load source must never silently delete or retarget send destinations.

The list shows small source/target counts so a layer with no routes or several destinations is easy to spot. It deliberately avoids a large checkbox matrix: connection × screen × layer combinations would quickly overwhelm that layout.

Send options per destination

Each destination has a collapsed Send options control with independent checkboxes for Position, Size, Crop, Fit mode, Mask, and Input. Position and size are separate so an operator can move a layer without intentionally replacing its dimensions. Crop includes edges, zoom and offsets; fit mode is separate from that framing. Input means the destination’s video source, not shared input processing settings.

The proposed default includes layout properties but leaves Input unchecked. Unchecked properties should keep their current device values; all unchecked should skip that destination. Following the load source should only link the destination address, not these checkboxes. Other destinations retain their own choices. In this mockup, controls only expand/collapse and toggle locally; nothing is saved or sent, and sample counts do not change.

Before implementation, resolve geometric dependencies: device crop, fit, size and position can affect one another. Preparation must use fresh device values for excluded properties and explain or reject combinations that cannot preserve them faithfully. It must not silently enable extra categories. The final transfer review should show included properties and explicitly flag input changes. This is a design proposal, not a promise of independent device fields.

Questions for visual review

  • Is the selected-layer approach comfortable, or is a project-wide route overview also needed? Start with this focused editor; an overview can remain a separate view.
  • Does “Follow load source” communicate the link clearly? A linked row should be read-only in the real UI until unlinked; native controls here are illustrative.
  • Should adding a destination open a searchable connection/screen/layer picker, then place the result in its connection group? This avoids three permanent selectors for connections that are already grouped.
  • Should mapping edits use Apply/Cancel as shown, or immediate undoable changes? Apply/Cancel is a proposal for review, not a change to current behavior.

States to design next, before implementation

Unmapped/send-only/load-only layers; unavailable connections; conflicting send destinations; different screen resolutions; large layer lists; narrow windows; keyboard navigation and long translations. A future connection selector needs explicit identities and reconnection behavior, not automatic matching by name.

A resolution mismatch should remain an actionable validation result until an explicit sizing policy is agreed. Do not invent automatic scaling in this window. Multi-device execution and partial failures belong to a later transfer review.

The existing stretch goal and mapping contract remain authoritative for implemented scope. This mockup changes neither.