windows-mcp 0.5.7__py3-none-any.whl → 0.5.9__py3-none-any.whl

windows_mcp/desktop/views.py

@@ -1,58 +1,60 @@
-from windows_mcp.tree.views import TreeState
-from dataclasses import dataclass
-from tabulate import tabulate
-from typing import Optional
-from PIL.Image import Image
-from enum import Enum
-
-class Browser(Enum):
-    CHROME='Chrome'
-    EDGE='Edge'
-    FIREFOX='Firefox'
-
-class Status(Enum):
-    MAXIMIZED='Maximized'
-    MINIMIZED='Minimized'
-    NORMAL='Normal'
-    HIDDEN='Hidden'
-
-
-@dataclass
-class App:
-    name:str
-    depth:int
-    status:Status
-    size:'Size'
-    handle: int
-    process_id:int
-    
-    def to_row(self):
-        return [self.name, self.depth, self.status.value, self.size.width, self.size.height, self.handle]
-
-@dataclass
-class Size:
-    width:int
-    height:int
-
-    def to_string(self):
-        return f'({self.width},{self.height})'
-
-@dataclass
-class DesktopState:
-    apps:list[App]
-    active_app:Optional[App]
-    screenshot:Image|None
-    tree_state:TreeState
-
-    def active_app_to_string(self):
-        if self.active_app is None:
-            return 'No active app found'
-        headers = ["Name", "Depth", "Status", "Width", "Height", "Handle"]
-        return tabulate([self.active_app.to_row()], headers=headers, tablefmt="simple")
-
-    def apps_to_string(self):
-        if not self.apps:
-            return 'No apps running in background'
-        headers = ["Name", "Depth", "Status", "Width", "Height", "Handle"]
-        rows = [app.to_row() for app in self.apps]
+from windows_mcp.tree.views import TreeState,BoundingBox
+from dataclasses import dataclass
+from tabulate import tabulate
+from typing import Optional
+from PIL.Image import Image
+from enum import Enum
+
+class Browser(Enum):
+    CHROME='Chrome'
+    EDGE='Edge'
+    FIREFOX='Firefox'
+
+class Status(Enum):
+    MAXIMIZED='Maximized'
+    MINIMIZED='Minimized'
+    NORMAL='Normal'
+    HIDDEN='Hidden'
+
+
+@dataclass
+class App:
+    name:str
+    runtime_id:tuple[int]
+    is_browser:bool
+    depth:int
+    status:Status
+    bounding_box:BoundingBox
+    handle: int
+    process_id:int
+    
+    def to_row(self):
+        return [self.name, self.depth, self.status.value, self.bounding_box.width, self.bounding_box.height, self.handle]
+
+@dataclass
+class Size:
+    width:int
+    height:int
+
+    def to_string(self):
+        return f'({self.width},{self.height})'
+
+@dataclass
+class DesktopState:
+    apps:list[App]
+    active_app:Optional[App]
+    screenshot:Optional[Image]=None
+    tree_state:Optional[TreeState]=None
+
+    def active_app_to_string(self):
+        if self.active_app is None:
+            return 'No active app found'
+        headers = ["Name", "Depth", "Status", "Width", "Height", "Handle"]
+        return tabulate([self.active_app.to_row()], headers=headers, tablefmt="simple")
+
+    def apps_to_string(self):
+        if not self.apps:
+            return 'No apps running in background'
+        headers = ["Name", "Depth", "Status", "Width", "Height", "Handle"]
+        rows = [app.to_row() for app in self.apps]
         return tabulate(rows, headers=headers, tablefmt="simple")

windows_mcp/tree/cache_utils.py

@@ -0,0 +1,126 @@
+"""
+UIA Caching Utilities for Performance Optimization
+
+This module provides utilities for implementing UI Automation caching
+to reduce cross-process COM calls during tree traversal.
+"""
+
+from windows_mcp.uia import CacheRequest, PropertyId, PatternId, TreeScope, Control
+from typing import Optional, Any
+import logging
+
+logger = logging.getLogger(__name__)
+
+
+class CacheRequestFactory:
+    """Factory for creating optimized cache requests for different scenarios."""
+    
+    @staticmethod
+    def create_tree_traversal_cache() -> CacheRequest:
+        """
+        Creates a cache request optimized for tree traversal.
+        Caches all commonly accessed properties and patterns.
+        
+        This cache request is designed to minimize COM calls during
+        the tree_traversal() operation in tree/service.py.
+        
+        Returns:
+            CacheRequest configured for tree traversal
+        """
+        cache_request = CacheRequest()
+        
+        # Set scope to cache element and its children
+        # This allows us to get children with pre-cached properties
+        cache_request.TreeScope = TreeScope.TreeScope_Element | TreeScope.TreeScope_Children
+        
+        # Basic identification properties
+        cache_request.AddProperty(PropertyId.NameProperty)
+        cache_request.AddProperty(PropertyId.AutomationIdProperty)
+        cache_request.AddProperty(PropertyId.LocalizedControlTypeProperty)
+        cache_request.AddProperty(PropertyId.AcceleratorKeyProperty)
+        cache_request.AddProperty(PropertyId.ClassNameProperty)
+        cache_request.AddProperty(PropertyId.ControlTypeProperty)
+        
+        # State properties for visibility and interaction checks
+        cache_request.AddProperty(PropertyId.IsEnabledProperty)
+        cache_request.AddProperty(PropertyId.IsOffscreenProperty)
+        cache_request.AddProperty(PropertyId.IsControlElementProperty)
+        cache_request.AddProperty(PropertyId.HasKeyboardFocusProperty)
+        cache_request.AddProperty(PropertyId.IsKeyboardFocusableProperty)
+        
+        # Layout properties
+        cache_request.AddProperty(PropertyId.BoundingRectangleProperty)
+        
+        # REMOVED: Expensive patterns and less critical properties to improve performance
+        # Patterns like LegacyIAccessible are very expensive to marshal for every element.
+        # We will fetch them live only for the few elements that actually need them.
+        
+        return cache_request
+    
+class CachedControlHelper:
+    """Helper class for working with cached controls."""
+    
+    @staticmethod
+    def build_cached_control(node: Control, cache_request: Optional[CacheRequest] = None) -> Control:
+        """
+        Build a cached version of a control.
+        
+        Args:
+            node: The control to cache
+            cache_request: Optional custom cache request. If None, uses tree traversal cache.
+        
+        Returns:
+            A control with cached properties, or the original control if caching fails
+        """
+        if cache_request is None:
+            cache_request = CacheRequestFactory.create_tree_traversal_cache()
+        
+        try:
+            cached_node = node.BuildUpdatedCache(cache_request)
+            cached_node._is_cached = True
+            return cached_node
+        except Exception as e:
+            logger.debug(f"Failed to build cached control: {e}")
+            return node
+    
+    @staticmethod
+    def get_cached_children(node: Control, cache_request: Optional[CacheRequest] = None) -> list[Control]:
+        """
+        Get children with pre-cached properties.
+        
+        This is the most significant optimization - it retrieves all children
+        with their properties already cached, eliminating the need for individual
+        property access calls on each child.
+        
+        Args:
+            node: The parent control
+            cache_request: Optional custom cache request. If None, uses tree traversal cache.
+        
+        Returns:
+            List of children with cached properties
+        """
+        if cache_request is None:
+            cache_request = CacheRequestFactory.create_tree_traversal_cache()
+        
+        # Ensure the cache request includes children
+        # Note: We do NOT set this here to avoid modifying shared cache request objects
+        # The caller is responsible for providing a CacheRequest with TreeScope_Children
+        if (cache_request.TreeScope & TreeScope.TreeScope_Children) == 0:
+             logger.warning("Cache request passed to get_cached_children does not have Children scope!")
+        
+        # Try to use existing cache first if available
+        try:
+            # Build updated cache that includes children
+            cached_node = node.BuildUpdatedCache(cache_request)
+            children = cached_node.GetCachedChildren()
+            
+            for child in children:
+                child._is_cached = True
+            
+            logger.debug(f"Retrieved {len(children)} cached children (newly built)")
+            return children
+            
+        except Exception as e:
+            logger.debug(f"Failed to get cached children, falling back to regular access: {e}")
+            return node.GetChildren()
+

windows_mcp/tree/config.py

@@ -1,51 +1,76 @@
-INTERACTIVE_CONTROL_TYPE_NAMES=set([
-    'ButtonControl',
-    'ListItemControl',
-    'MenuItemControl',
-    'EditControl',
-    'CheckBoxControl',
-    'RadioButtonControl',
-    'ComboBoxControl',
-    'HyperlinkControl',
-    'SplitButtonControl',
-    'TabItemControl',
-    'TreeItemControl',
-    'DataItemControl',
-    'HeaderItemControl',
-    'TextBoxControl',
-    'SpinnerControl',
-    'ScrollBarControl'
-])
-
-DOCUMENT_CONTROL_TYPE_NAMES=set([
-    'DocumentControl'
-])
-
-STRUCTURAL_CONTROL_TYPE_NAMES = set([
-    'PaneControl',
-    'GroupControl',
-    'CustomControl'
-])
-
-INFORMATIVE_CONTROL_TYPE_NAMES=set([
-    'TextControl',
-    'ImageControl',
-    'StatusBarControl',
-    # 'ProgressBarControl',
-    # 'ToolTipControl',
-    # 'TitleBarControl',
-    # 'SeparatorControl',
-    # 'HeaderControl',
-    # 'HeaderItemControl',
-])
-
-DEFAULT_ACTIONS=set([
-    'Click',
-    'Press',
-    'Jump',
-    'Check',
-    'Uncheck',
-    'Double Click'
-])
-
+INTERACTIVE_CONTROL_TYPE_NAMES=set([
+    'ButtonControl',
+    'ListItemControl',
+    'MenuItemControl',
+    'EditControl',
+    'CheckBoxControl',
+    'RadioButtonControl',
+    'ComboBoxControl',
+    'HyperlinkControl',
+    'SplitButtonControl',
+    'TabItemControl',
+    'TreeItemControl',
+    'DataItemControl',
+    'HeaderItemControl',
+    'TextBoxControl',
+    'SpinnerControl',
+    'ScrollBarControl'
+])
+
+INTERACTIVE_ROLES = {
+    # Buttons
+    'PushButton', 'SplitButton', 'ButtonDropDown', 'ButtonMenu', 
+    'ButtonDropDownGrid', 'OutlineButton',
+    
+    # Links
+    'Link',
+    
+    # Inputs & Selection
+    'Text', 'IpAddress', 'HotkeyField', 'ComboBox', 'DropList',
+    'CheckButton', 'RadioButton',
+    
+    # Menus & Tabs
+    'MenuItem', 'ListItem', 'PageTab',
+    
+    # Trees
+    'OutlineItem',
+    
+    # Values
+    'Slider', 'SpinButton', 'Dial', 'ScrollBar', 'Grip',
+    
+    # Grids
+    'ColumnHeader', 'RowHeader', 'Cell'
+}
+
+DOCUMENT_CONTROL_TYPE_NAMES=set([
+    'DocumentControl'
+])
+
+STRUCTURAL_CONTROL_TYPE_NAMES = set([
+    'PaneControl',
+    'GroupControl',
+    'CustomControl'
+])
+
+INFORMATIVE_CONTROL_TYPE_NAMES=set([
+    'TextControl',
+    'ImageControl',
+    'StatusBarControl',
+    # 'ProgressBarControl',
+    # 'ToolTipControl',
+    # 'TitleBarControl',
+    # 'SeparatorControl',
+    # 'HeaderControl',
+    # 'HeaderItemControl',
+])
+
+DEFAULT_ACTIONS=set([
+    'Click',
+    'Press',
+    'Jump',
+    'Check',
+    'Uncheck',
+    'Double Click'
+])
+
 THREAD_MAX_RETRIES = 3