[kernel-847] add functionality to get cursor position via API

server/cmd/api/api/computer.go

@@ -9,6 +9,7 @@ import (
 	"os"
 	"os/exec"
 	"strconv"
+	"strings"
 	"syscall"
 	"time"
 
@@ -426,6 +427,85 @@ func (s *ApiService) SetCursor(ctx context.Context, request oapi.SetCursorReques
 	return oapi.SetCursor200JSONResponse{Ok: true}, nil
 }
 
+// parseMousePosition parses xdotool getmouselocation --shell output.
+// Expected format:
+//
+//	X=100
+//	Y=200
+//	SCREEN=0
+//	WINDOW=12345
+//
+// Returns x, y coordinates and an error if parsing fails.
+func parseMousePosition(output string) (x, y int, err error) {
+	outStr := strings.TrimSpace(output)
+	if outStr == "" {
+		return 0, 0, fmt.Errorf("empty output")
+	}
+
+	var xParsed, yParsed bool
+
+	for line := range strings.SplitSeq(outStr, "\n") {
+		line = strings.TrimSpace(line)
+		parts := strings.SplitN(line, "=", 2)
+		if len(parts) != 2 {
+			continue
+		}
+		key, value := parts[0], parts[1]
+		switch key {
+		case "X":
+			if parsed, parseErr := strconv.Atoi(value); parseErr == nil {
+				x = parsed
+				xParsed = true
+			}
+		case "Y":
+			if parsed, parseErr := strconv.Atoi(value); parseErr == nil {
+				y = parsed
+				yParsed = true
+			}
+		}
+		// Early exit once both coordinates are found
+		if xParsed && yParsed {
+			break
+		}
+	}
+
+	if !xParsed || !yParsed {
+		return 0, 0, fmt.Errorf("failed to parse coordinates from output: %q", outStr)
+	}
+
+	return x, y, nil
+}
+
+func (s *ApiService) GetMousePosition(ctx context.Context, request oapi.GetMousePositionRequestObject) (oapi.GetMousePositionResponseObject, error) {
+	log := logger.FromContext(ctx)
+
+	// serialize input operations to avoid race conditions with other xdotool commands
+	s.inputMu.Lock()
+	defer s.inputMu.Unlock()
+
+	// Execute xdotool getmouselocation --shell for parseable output
+	output, err := defaultXdoTool.Run(ctx, "getmouselocation", "--shell")
+	if err != nil {
+		log.Error("xdotool getmouselocation failed", "err", err, "output", string(output))
+		return oapi.GetMousePosition500JSONResponse{InternalErrorJSONResponse: oapi.InternalErrorJSONResponse{
+			Message: "failed to get mouse position"},
+		}, nil
+	}
+
+	x, y, err := parseMousePosition(string(output))
+	if err != nil {
+		log.Error("failed to parse mouse position", "err", err, "output", string(output))
+		return oapi.GetMousePosition500JSONResponse{InternalErrorJSONResponse: oapi.InternalErrorJSONResponse{
+			Message: "failed to parse mouse position from xdotool output"},
+		}, nil
+	}
+
+	return oapi.GetMousePosition200JSONResponse{
+		X: x,
+		Y: y,
+	}, nil
+}
+
 func (s *ApiService) PressKey(ctx context.Context, request oapi.PressKeyRequestObject) (oapi.PressKeyResponseObject, error) {
 	log := logger.FromContext(ctx)
 

server/cmd/api/api/computer_test.go

@@ -49,3 +49,94 @@ func TestGenerateRelativeSteps_DiagonalsAndSlopes(t *testing.T) {
 		require.Equal(t, 5, countSteps(steps), "count mismatch")
 	}
 }
+
+// TestParseMousePosition tests the parseMousePosition helper function
+func TestParseMousePosition(t *testing.T) {
+	tests := []struct {
+		name        string
+		output      string
+		expectX     int
+		expectY     int
+		expectError bool
+	}{
+		{
+			name:        "valid output",
+			output:      "X=100\nY=200\nSCREEN=0\nWINDOW=12345\n",
+			expectX:     100,
+			expectY:     200,
+			expectError: false,
+		},
+		{
+			name:        "valid output with extra whitespace",
+			output:      "  X=512  \n  Y=384  \n  SCREEN=0  \n  WINDOW=67890  \n",
+			expectX:     512,
+			expectY:     384,
+			expectError: false,
+		},
+		{
+			name:        "missing Y coordinate",
+			output:      "X=100\nSCREEN=0\nWINDOW=12345\n",
+			expectError: true,
+		},
+		{
+			name:        "missing X coordinate",
+			output:      "Y=200\nSCREEN=0\nWINDOW=12345\n",
+			expectError: true,
+		},
+		{
+			name:        "empty output",
+			output:      "",
+			expectError: true,
+		},
+		{
+			name:        "whitespace only",
+			output:      "   \n  \t  \n",
+			expectError: true,
+		},
+		{
+			name:        "non-numeric X value",
+			output:      "X=abc\nY=200\nSCREEN=0\nWINDOW=12345\n",
+			expectError: true,
+		},
+		{
+			name:        "non-numeric Y value",
+			output:      "X=100\nY=xyz\nSCREEN=0\nWINDOW=12345\n",
+			expectError: true,
+		},
+		{
+			name:        "zero coordinates",
+			output:      "X=0\nY=0\nSCREEN=0\nWINDOW=12345\n",
+			expectX:     0,
+			expectY:     0,
+			expectError: false,
+		},
+		{
+			name:        "negative coordinates",
+			output:      "X=-50\nY=-100\nSCREEN=0\nWINDOW=12345\n",
+			expectX:     -50,
+			expectY:     -100,
+			expectError: false,
+		},
+		{
+			name:        "large coordinates",
+			output:      "X=3840\nY=2160\nSCREEN=0\nWINDOW=12345\n",
+			expectX:     3840,
+			expectY:     2160,
+			expectError: false,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			x, y, err := parseMousePosition(tt.output)
+
+			if tt.expectError {
+				require.Error(t, err, "expected parsing to fail")
+			} else {
+				require.NoError(t, err, "expected successful parsing")
+				require.Equal(t, tt.expectX, x, "X coordinate mismatch")
+				require.Equal(t, tt.expectY, y, "Y coordinate mismatch")
+			}
+		})
+	}
+}