feat: add html5drag action and multi-strategy drag support

Adds html5drag action that dispatches synthetic HTML5 DnD events (dragstart,
dragenter, dragover, drop, dragend) via DOM.resolveNode + Runtime.evaluate.
Uses data attributes to bridge CDP node IDs to JS DOM elements. Adds
TargetRef/TargetNodeID fields to ActionRequest for drop target specification.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: fry-lobster

internal/bridge/action_pointer.go

@@ -132,6 +132,29 @@ func (b *Bridge) actionDrag(ctx context.Context, req ActionRequest) (map[string]
 	return nil, fmt.Errorf("need selector, ref, or nodeId")
 }
 
+func (b *Bridge) actionHTML5Drag(ctx context.Context, req ActionRequest) (map[string]any, error) {
+	if req.TargetNodeID == 0 {
+		return nil, fmt.Errorf("targetNodeId required for html5drag")
+	}
+	var sourceNodeID int64
+	if req.NodeID > 0 {
+		sourceNodeID = req.NodeID
+	} else if req.Selector != "" {
+		node, err := firstNodeBySelector(ctx, req.Selector)
+		if err != nil {
+			return nil, err
+		}
+		sourceNodeID = int64(node.BackendNodeID)
+	} else {
+		return nil, fmt.Errorf("need selector, ref, or nodeId for source element")
+	}
+	err := HTML5DragByNodeID(ctx, sourceNodeID, req.TargetNodeID)
+	if err != nil {
+		return nil, err
+	}
+	return map[string]any{"dragged": true, "strategy": "html5"}, nil
+}
+
 func (b *Bridge) actionHumanClick(ctx context.Context, req ActionRequest) (map[string]any, error) {
 	if req.NodeID > 0 {
 		// req.NodeID is a backendDOMNodeId from the accessibility tree

internal/bridge/action_registry.go

@@ -21,6 +21,7 @@ const (
 	ActionKeyUp          = "keyup"
 	ActionMouseDown      = "mousedown"
 	ActionMouseUp        = "mouseup"
+	ActionHTML5Drag      = "html5drag"
 	ActionScrollIntoView = "scrollintoview"
 )
 
@@ -46,6 +47,7 @@ func (b *Bridge) InitActionRegistry() {
 		ActionKeyUp:          b.actionKeyUp,
 		ActionMouseDown:      b.actionMouseDown,
 		ActionMouseUp:        b.actionMouseUp,
+		ActionHTML5Drag:      b.actionHTML5Drag,
 		ActionScrollIntoView: b.actionScrollIntoView,
 	}
 }

internal/bridge/bridge.go

@@ -386,8 +386,10 @@ type ActionRequest struct {
 
 	ScrollX int `json:"scrollX"`
 	ScrollY int `json:"scrollY"`
-	DragX   int `json:"dragX"`
-	DragY   int `json:"dragY"`
+	DragX        int   `json:"dragX"`
+	DragY        int   `json:"dragY"`
+	TargetRef    string `json:"targetRef,omitempty"`
+	TargetNodeID int64  `json:"targetNodeId,omitempty"`
 
 	WaitNav bool   `json:"waitNav"`
 	Fast    bool   `json:"fast"`

internal/bridge/cdpops/pointer.go

@@ -257,6 +257,90 @@ func MouseUpByNodeID(ctx context.Context, nodeID int64) error {
 	}))
 }
 
+// HTML5DragByNodeID performs a synthetic HTML5 Drag and Drop between two elements.
+// Uses JS to dispatch dragstart, dragover, drop, and dragend events.
+func HTML5DragByNodeID(ctx context.Context, sourceNodeID, targetNodeID int64) error {
+	js := `
+(function(srcId, tgtId) {
+  function nodeById(id) {
+    return document.querySelector('[data-ptab-nid="' + id + '"]') ||
+           (function() {
+             const all = document.querySelectorAll('*');
+             for (const el of all) if (el.__backendNodeId === id) return el;
+             return null;
+           })();
+  }
+  // Resolve via CDP-injected attribute or fallback
+  const src = nodeById(srcId);
+  const tgt = nodeById(tgtId);
+  if (!src || !tgt) return JSON.stringify({error: 'element not found', src: !!src, tgt: !!tgt});
+
+  const dt = new DataTransfer();
+  const opts = {bubbles: true, cancelable: true, dataTransfer: dt};
+  src.dispatchEvent(new DragEvent('dragstart', opts));
+  tgt.dispatchEvent(new DragEvent('dragenter', opts));
+  tgt.dispatchEvent(new DragEvent('dragover', opts));
+  tgt.dispatchEvent(new DragEvent('drop', opts));
+  src.dispatchEvent(new DragEvent('dragend', opts));
+  return JSON.stringify({ok: true});
+})
+`
+	// We need to resolve backendNodeId to DOM elements. Use DOM.resolveNode + evaluate.
+	return chromedp.Run(ctx,
+		// First, tag both elements with a data attribute so JS can find them
+		chromedp.ActionFunc(func(ctx context.Context) error {
+			js := fmt.Sprintf(`document.querySelector('[data-ptab-nid="%d"]')?.removeAttribute('data-ptab-nid'); void 0`, sourceNodeID)
+			return chromedp.FromContext(ctx).Target.Execute(ctx, "Runtime.evaluate", map[string]any{"expression": js}, nil)
+		}),
+		chromedp.ActionFunc(func(ctx context.Context) error {
+			// Resolve source node and tag it
+			var result map[string]any
+			if err := chromedp.FromContext(ctx).Target.Execute(ctx, "DOM.resolveNode", map[string]any{
+				"backendNodeId": sourceNodeID,
+			}, &result); err != nil {
+				return fmt.Errorf("resolve source: %w", err)
+			}
+			objectID, _ := result["object"].(map[string]any)["objectId"].(string)
+			if objectID == "" {
+				return fmt.Errorf("could not resolve source node %d", sourceNodeID)
+			}
+			return chromedp.FromContext(ctx).Target.Execute(ctx, "Runtime.callFunctionOn", map[string]any{
+				"objectId":            objectID,
+				"functionDeclaration": fmt.Sprintf(`function() { this.setAttribute('data-ptab-nid', '%d'); }`, sourceNodeID),
+			}, nil)
+		}),
+		chromedp.ActionFunc(func(ctx context.Context) error {
+			// Resolve target node and tag it
+			var result map[string]any
+			if err := chromedp.FromContext(ctx).Target.Execute(ctx, "DOM.resolveNode", map[string]any{
+				"backendNodeId": targetNodeID,
+			}, &result); err != nil {
+				return fmt.Errorf("resolve target: %w", err)
+			}
+			objectID, _ := result["object"].(map[string]any)["objectId"].(string)
+			if objectID == "" {
+				return fmt.Errorf("could not resolve target node %d", targetNodeID)
+			}
+			return chromedp.FromContext(ctx).Target.Execute(ctx, "Runtime.callFunctionOn", map[string]any{
+				"objectId":            objectID,
+				"functionDeclaration": fmt.Sprintf(`function() { this.setAttribute('data-ptab-nid', '%d'); }`, targetNodeID),
+			}, nil)
+		}),
+		chromedp.ActionFunc(func(ctx context.Context) error {
+			expr := fmt.Sprintf(`%s(%d, %d)`, js, sourceNodeID, targetNodeID)
+			return chromedp.FromContext(ctx).Target.Execute(ctx, "Runtime.evaluate", map[string]any{"expression": expr}, nil)
+		}),
+		// Clean up tags
+		chromedp.ActionFunc(func(ctx context.Context) error {
+			cleanup := fmt.Sprintf(`document.querySelectorAll('[data-ptab-nid]').forEach(el => el.removeAttribute('data-ptab-nid')); void 0`)
+			_ = cleanup
+			return chromedp.FromContext(ctx).Target.Execute(ctx, "Runtime.evaluate", map[string]any{
+				"expression": `document.querySelectorAll('[data-ptab-nid]').forEach(el => el.removeAttribute('data-ptab-nid')); void 0`,
+			}, nil)
+		}),
+	)
+}
+
 func HoverByCoordinate(ctx context.Context, x, y float64) error {
 	if x < 0 || y < 0 {
 		return fmt.Errorf("x/y coordinates must be >= 0")

internal/bridge/cdpops_facade.go

@@ -63,6 +63,10 @@ func DragByNodeID(ctx context.Context, nodeID int64, dx, dy int) error {
 	return bridgecdpops.DragByNodeID(ctx, nodeID, dx, dy)
 }
 
+func HTML5DragByNodeID(ctx context.Context, sourceNodeID, targetNodeID int64) error {
+	return bridgecdpops.HTML5DragByNodeID(ctx, sourceNodeID, targetNodeID)
+}
+
 func MouseDownByCoordinate(ctx context.Context, x, y float64) error {
 	return bridgecdpops.MouseDownByCoordinate(ctx, x, y)
 }