kailash 0.9.2__py3-none-any.whl → 0.9.4__py3-none-any.whl

kailash/runtime/parallel.py

@@ -199,7 +199,7 @@ class ParallelRuntime:
         )
 
         # Process nodes until all are complete
-        while ready_nodes or pending_nodes:
+        while ready_nodes or pending_nodes or node_tasks:
             # Schedule ready nodes up to max_workers limit
             while ready_nodes and len(node_tasks) < self.max_workers:
                 node_id = ready_nodes.popleft()

kailash/runtime/performance_monitor.py

@@ -0,0 +1,215 @@
+"""Performance monitoring for conditional execution.
+
+This module provides performance tracking and automatic fallback capabilities
+for the conditional execution feature.
+"""
+
+import logging
+import time
+from collections import deque
+from dataclasses import dataclass, field
+from typing import Dict, Optional, Tuple
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class ExecutionMetrics:
+    """Metrics for a single execution."""
+
+    execution_time: float
+    node_count: int
+    skipped_nodes: int
+    memory_usage: Optional[float] = None
+    execution_mode: str = "route_data"
+
+    @property
+    def nodes_per_second(self) -> float:
+        """Calculate execution rate."""
+        if self.execution_time > 0:
+            return self.node_count / self.execution_time
+        return 0.0
+
+    @property
+    def skip_ratio(self) -> float:
+        """Calculate percentage of nodes skipped."""
+        total = self.node_count + self.skipped_nodes
+        if total > 0:
+            return self.skipped_nodes / total
+        return 0.0
+
+
+class PerformanceMonitor:
+    """Monitor performance and make mode switching decisions.
+
+    Tracks execution performance of conditional vs standard execution
+    and automatically switches modes based on performance thresholds.
+    """
+
+    def __init__(
+        self,
+        performance_threshold: float = 0.9,  # Switch if conditional is 90% slower
+        sample_size: int = 10,
+        min_samples: int = 3,
+    ):
+        """Initialize performance monitor.
+
+        Args:
+            performance_threshold: Ratio threshold for switching modes (0.9 = 10% slower triggers switch)
+            sample_size: Number of recent executions to track
+            min_samples: Minimum samples before making switching decisions
+        """
+        self.performance_threshold = performance_threshold
+        self.sample_size = sample_size
+        self.min_samples = min_samples
+
+        # Track metrics for each mode
+        self.metrics: Dict[str, deque] = {
+            "route_data": deque(maxlen=sample_size),
+            "skip_branches": deque(maxlen=sample_size),
+        }
+
+        # Performance statistics
+        self.mode_performance: Dict[str, float] = {
+            "route_data": 0.0,
+            "skip_branches": 0.0,
+        }
+
+        # Current recommendation
+        self.recommended_mode = "route_data"  # Safe default
+        self._last_evaluation_time = 0.0
+        self._evaluation_interval = 60.0  # Re-evaluate every minute
+
+    def record_execution(self, metrics: ExecutionMetrics) -> None:
+        """Record execution metrics.
+
+        Args:
+            metrics: Execution metrics to record
+        """
+        mode = metrics.execution_mode
+        if mode in self.metrics:
+            self.metrics[mode].append(metrics)
+            logger.debug(
+                f"Recorded {mode} execution: {metrics.execution_time:.3f}s, "
+                f"{metrics.node_count} nodes, {metrics.skipped_nodes} skipped"
+            )
+
+    def should_switch_mode(self, current_mode: str) -> Tuple[bool, str, str]:
+        """Determine if mode should be switched based on performance.
+
+        Args:
+            current_mode: Currently active execution mode
+
+        Returns:
+            Tuple of (should_switch, recommended_mode, reason)
+        """
+        # Check if enough time has passed since last evaluation
+        current_time = time.time()
+        if current_time - self._last_evaluation_time < self._evaluation_interval:
+            return False, current_mode, "Too soon since last evaluation"
+
+        self._last_evaluation_time = current_time
+
+        # Calculate average performance for each mode
+        route_data_avg = self._calculate_average_performance("route_data")
+        skip_branches_avg = self._calculate_average_performance("skip_branches")
+
+        # Not enough data to make decision
+        if route_data_avg is None or skip_branches_avg is None:
+            return False, current_mode, "Insufficient performance data"
+
+        # Update performance statistics
+        self.mode_performance["route_data"] = route_data_avg
+        self.mode_performance["skip_branches"] = skip_branches_avg
+
+        # Determine recommendation based on performance
+        if skip_branches_avg < route_data_avg * self.performance_threshold:
+            # skip_branches is significantly faster
+            self.recommended_mode = "skip_branches"
+            if current_mode != "skip_branches":
+                reason = (
+                    f"skip_branches mode is {(1 - skip_branches_avg/route_data_avg)*100:.1f}% faster "
+                    f"({skip_branches_avg:.3f}s vs {route_data_avg:.3f}s)"
+                )
+                return True, "skip_branches", reason
+        else:
+            # route_data is faster or difference is negligible
+            self.recommended_mode = "route_data"
+            if current_mode != "route_data":
+                reason = (
+                    f"route_data mode is faster or difference negligible "
+                    f"({route_data_avg:.3f}s vs {skip_branches_avg:.3f}s)"
+                )
+                return True, "route_data", reason
+
+        return False, current_mode, "Current mode is optimal"
+
+    def _calculate_average_performance(self, mode: str) -> Optional[float]:
+        """Calculate average execution time for a mode.
+
+        Args:
+            mode: Execution mode to analyze
+
+        Returns:
+            Average execution time per node, or None if insufficient data
+        """
+        if mode not in self.metrics:
+            return None
+
+        metrics_list = list(self.metrics[mode])
+        if len(metrics_list) < self.min_samples:
+            return None
+
+        # Calculate average time per node
+        total_time = sum(m.execution_time for m in metrics_list)
+        total_nodes = sum(m.node_count for m in metrics_list)
+
+        if total_nodes > 0:
+            return total_time / total_nodes
+        return None
+
+    def get_performance_report(self) -> Dict[str, any]:
+        """Generate performance report.
+
+        Returns:
+            Dictionary with performance statistics
+        """
+        report = {
+            "recommended_mode": self.recommended_mode,
+            "mode_performance": self.mode_performance.copy(),
+            "sample_counts": {
+                mode: len(metrics) for mode, metrics in self.metrics.items()
+            },
+            "performance_threshold": self.performance_threshold,
+        }
+
+        # Add detailed metrics if available
+        for mode, metrics_deque in self.metrics.items():
+            if metrics_deque:
+                metrics_list = list(metrics_deque)
+                report[f"{mode}_stats"] = {
+                    "avg_execution_time": sum(m.execution_time for m in metrics_list)
+                    / len(metrics_list),
+                    "avg_nodes": sum(m.node_count for m in metrics_list)
+                    / len(metrics_list),
+                    "avg_skip_ratio": sum(m.skip_ratio for m in metrics_list)
+                    / len(metrics_list),
+                    "total_executions": len(metrics_list),
+                }
+
+        return report
+
+    def clear_metrics(self, mode: Optional[str] = None) -> None:
+        """Clear performance metrics.
+
+        Args:
+            mode: Specific mode to clear, or None to clear all
+        """
+        if mode:
+            if mode in self.metrics:
+                self.metrics[mode].clear()
+        else:
+            for m in self.metrics.values():
+                m.clear()
+
+        logger.info(f"Cleared performance metrics for: {mode or 'all modes'}")

kailash/runtime/validation/import_validator.py

@@ -207,6 +207,7 @@ class ImportPathValidator:
         - Module exists as sibling to current file
         - Module is not a known SDK module
         - Module is not a standard library module
+        - Module is not a legitimate top-level package
         """
         # Skip if it's a known SDK module
         if module_name in self.sdk_modules:
@@ -228,6 +229,12 @@ class ImportPathValidator:
         ]:
             return False
 
+        # Skip common top-level package names that are meant for absolute imports
+        # These are legitimate when used as project structure roots
+        top_level_packages = ["src", "lib", "app", "pkg"]
+        if module_name in top_level_packages:
+            return False
+
         # Check if module exists as sibling
         parent_dir = file_path.parent
         possible_module = parent_dir / module_name