wasm4pm 26.6.12

#!/usr/bin/env python3
"""
Generate pm4py reference outputs for comprehensive parity testing.

This script uses pm4py to discover models and generate reference outputs
that can be compared against wasm4pm's outputs.

Usage:
    cd /Users/sac/chatmangpt/wasm4pm
    python3 wasm4pm/tests/generate_pm4py_reference.py
"""

import pm4py
import json
from pathlib import Path
from pm4py.algo.evaluation.generalization import algorithm as generalization
from pm4py.algo.evaluation.simplicity import algorithm as simplicity

# Fixtures directory
FIXTURES_DIR = Path("wasm4pm/tests/fixtures")
OUTPUT_DIR = FIXTURES_DIR

def load_running_example():
    """Load the canonical running-example.xes log."""
    xes_path = FIXTURES_DIR / "running-example.xes"
    return pm4py.read_xes(str(xes_path))

def save_event_log_json(log, output_path):
    """Save event log as JSON for wasm4pm to load."""
    output = {
        "attributes": {},
        "traces": []
    }

    # Check if log is a DataFrame (pm4py default) or EventLog
    if hasattr(log, 'iterrows'):
        # DataFrame format
        from collections import defaultdict

        traces = defaultdict(list)
        for idx, row in log.iterrows():
            case_id = str(row.get('case:concept:name', idx))
            event = {"attributes": {}}

            for col in row.index:
                if col == 'case:concept:name':
                    continue
                value = row[col]
                if col == 'concept:name':
                    event["attributes"]["activity"] = {"tag": "String", "value": str(value)}
                elif col == 'time:timestamp':
                    if hasattr(value, 'isoformat'):
                        event["attributes"]["timestamp"] = {"tag": "Date", "value": value.isoformat()}
                    else:
                        event["attributes"]["timestamp"] = {"tag": "String", "value": str(value)}
                else:
                    if value is None:
                        event["attributes"][col] = None
                    elif isinstance(value, str):
                        event["attributes"][col] = {"tag": "String", "value": value}
                    elif isinstance(value, int):
                        event["attributes"][col] = {"tag": "Int", "value": value}
                    elif isinstance(value, float):
                        event["attributes"][col] = {"tag": "Float", "value": value}
                    elif isinstance(value, bool):
                        event["attributes"][col] = {"tag": "Boolean", "value": value}
                    else:
                        event["attributes"][col] = {"tag": "String", "value": str(value)}

            traces[case_id].append(event)

        for case_id, events in traces.items():
            trace_data = {
                "attributes": {"case:concept:name": {"tag": "String", "value": case_id}},
                "events": events
            }
            output["traces"].append(trace_data)

    else:
        # EventLog format (traditional pm4py)
        for case_id, trace in enumerate(log, start=1):
            trace_data = {
                "attributes": {"case:concept:name": {"tag": "String", "value": str(case_id)}},
                "events": []
            }

            for event in trace:
                event_data = {
                    "attributes": {}
                }
                for key in event:
                    value = event[key]
                    if key == "concept:name":
                        event_data["attributes"]["activity"] = {"tag": "String", "value": str(value)}
                    elif key == "time:timestamp" and hasattr(value, 'isoformat'):
                        event_data["attributes"]["timestamp"] = {"tag": "Date", "value": value.isoformat()}
                    else:
                        if value is None:
                            event_data["attributes"][key] = None
                        elif isinstance(value, str):
                            event_data["attributes"][key] = {"tag": "String", "value": value}
                        elif isinstance(value, int):
                            event_data["attributes"][key] = {"tag": "Int", "value": value}
                        elif isinstance(value, float):
                            event_data["attributes"][key] = {"tag": "Float", "value": value}
                        elif isinstance(value, bool):
                            event_data["attributes"][key] = {"tag": "Boolean", "value": value}
                        else:
                            event_data["attributes"][key] = {"tag": "String", "value": str(value)}
                trace_data["events"].append(event_data)

            output["traces"].append(trace_data)

    with open(output_path, 'w') as f:
        json.dump(output, f, indent=2)
    print(f"Event log JSON saved to: {output_path}")

def discover_and_save_dfg(log, output_path):
    """Discover DFG using pm4py and save to JSON."""
    dfg, start_activities, end_activities = pm4py.discover_dfg(log)

    # Get all unique activities from the DFG
    all_activities = set()
    for (a, b), count in dfg.items():
        all_activities.add(a)
        all_activities.add(b)

    # Convert to wasm4pm-compatible format
    output = {
        "activities": sorted(list(all_activities)),
        "edges": [
            {"from": a, "to": b, "frequency": count}
            for (a, b), count in dfg.items()
        ],
        "start_activities": [
            {"activity": act, "count": count}
            for act, count in sorted(start_activities.items())
        ],
        "end_activities": [
            {"activity": act, "count": count}
            for act, count in sorted(end_activities.items())
        ],
    }

    with open(output_path, 'w') as f:
        json.dump(output, f, indent=2)
    print(f"DFG output saved to: {output_path}")

def discover_and_save_inductive_miner(log, output_path):
    """Discover process tree using inductive miner and save to JSON."""
    tree = pm4py.discover_process_tree_inductive(log)

    # Convert to JSON-serializable format
    def tree_to_dict(tree):
        if tree is None:
            return None
        result = {
            "label": tree.label if hasattr(tree, 'label') else None,
            "operator": str(tree.operator) if hasattr(tree, 'operator') else None,
            "children": []
        }
        if hasattr(tree, 'children') and tree.children:
            result["children"] = [tree_to_dict(child) for child in tree.children]
        return result

    output = {
        "tree": tree_to_dict(tree),
        "activities": sorted(list(pm4py.get_event_attributes(log))),
    }

    with open(output_path, 'w') as f:
        json.dump(output, f, indent=2, default=str)
    print(f"Inductive miner output saved to: {output_path}")

def discover_and_save_alpha_miner(log, output_path):
    """Discover Petri net using alpha miner and save to JSON."""
    net, im, fm = pm4py.discover_petri_net_alpha(log)

    # Convert to JSON-serializable format
    output = {
        "places": [
            {
                "name": place.name,
                "label": place.properties.get("label", "") if hasattr(place, 'properties') and place.properties else ""
            }
            for place in net.places
        ],
        "transitions": [
            {
                "name": trans.name,
                "label": trans.label if hasattr(trans, 'label') else None,
                "silent": trans.label is None if hasattr(trans, 'label') else False
            }
            for trans in net.transitions
        ],
        "arcs": [
            {
                "source": arc.source.name,
                "target": arc.target.name,
                "weight": arc.weight if hasattr(arc, 'weight') else 1
            }
            for arc in net.arcs
        ],
        "initial_marking": {place.name: count for place, count in im.items()},
        "final_marking": {place.name: count for place, count in fm.items()},
    }

    with open(output_path, 'w') as f:
        json.dump(output, f, indent=2)
    print(f"Alpha miner output saved to: {output_path}")

def discover_and_save_heuristic_miner(log, output_path):
    """Discover Petri net using heuristic miner and save to JSON."""
    from pm4py.algo.discovery.heuristics import algorithm as heuristics_miner
    heu_params = heuristics_miner.Variants.CLASSIC.value.Parameters
    net, im, fm = heuristics_miner.apply(log, parameters={
        heu_params.DEPENDENCY_THRESH: 0.5
    })

    # Convert to JSON-serializable format
    output = {
        "places": [
            {
                "name": place.name,
                "label": place.properties.get("label", "") if hasattr(place, 'properties') and place.properties else ""
            }
            for place in net.places
        ],
        "transitions": [
            {
                "name": trans.name,
                "label": trans.label if hasattr(trans, 'label') else None,
                "silent": trans.label is None if hasattr(trans, 'label') else False
            }
            for trans in net.transitions
        ],
        "arcs": [
            {
                "source": arc.source.name,
                "target": arc.target.name,
                "weight": arc.weight if hasattr(arc, 'weight') else 1
            }
            for arc in net.arcs
        ],
        "initial_marking": {place.name: count for place, count in im.items()},
        "final_marking": {place.name: count for place, count in fm.items()},
    }

    with open(output_path, 'w') as f:
        json.dump(output, f, indent=2)
    print(f"Heuristic miner output saved to: {output_path}")

def save_token_replay_conformance(log, output_path):
    """Calculate token replay conformance and save to JSON."""
    # Discover a process tree using inductive miner
    tree = pm4py.discover_process_tree_inductive(log)

    # Convert to Petri net for conformance checking
    from pm4py.convert import convert_to_petri_net
    net, im, fm = convert_to_petri_net(tree)

    # Calculate token replay fitness
    from pm4py.algo.conformance.tokenreplay import algorithm as token_replay
    replay_results = token_replay.apply(log, net, im, fm)

    # Aggregate results
    total_traces = len(replay_results)
    fitted_traces = sum(1 for res in replay_results if res["trace_is_fit"])
    avg_trace_fitness = sum(res["trace_fitness"] for res in replay_results) / total_traces if total_traces > 0 else 0.0

    output = {
        "total_traces": total_traces,
        "fitted_traces": fitted_traces,
        "unfit_traces": total_traces - fitted_traces,
        "avg_trace_fitness": avg_trace_fitness,
        "fitness_percentage": (fitted_traces / total_traces * 100) if total_traces > 0 else 0.0,
    }

    with open(output_path, 'w') as f:
        json.dump(output, f, indent=2)
    print(f"Token replay conformance output saved to: {output_path}")

def save_footprints(log, output_path):
    """Calculate footprints and save to JSON."""
    try:
        from pm4py.algo.discovery.footprints import algorithm as footprints_discovery
        footprints_result = footprints_discovery.apply(log, variant=footprints_discovery.Variants.CLASSIC)

        output = {
            "footprints": str(footprints_result),
        }
    except Exception as e:
        # Footprints API may vary, save a simple version
        dfg, start, end = pm4py.discover_dfg(log)
        output = {
            "footprints": {
                "start_activities": list(start.keys()),
                "end_activities": list(end.keys()),
                "dfg_edges": list(dfg.keys()),
            }
        }

    with open(output_path, 'w') as f:
        json.dump(output, f, indent=2)
    print(f"Footprints output saved to: {output_path}")

def save_model_quality_metrics(log, output_path):
    """Calculate model quality metrics and save to JSON."""
    # Discover process tree
    tree = pm4py.discover_process_tree_inductive(log)

    try:
        # Calculate quality metrics
        from pm4py.algo.evaluation.replay_fitness import algorithm as fitness_evaluator
        from pm4py.algo.evaluation.precision import algorithm as precision_evaluator

        fitness_result = fitness_evaluator.apply(log, tree, variant=fitness_evaluator.Variants.TOKEN_BASED)
        precision_result = precision_evaluator.apply(log, tree, variant=precision_evaluator.Variants.ETCONFORMANCE_TOKEN)

        output = {
            "fitness": {
                "average_trace_fitness": fitness_result["average_trace_fitness"] if isinstance(fitness_result, dict) else fitness_result,
                "perc_fit_traces": fitness_result.get("perc_fit_traces", 100) if isinstance(fitness_result, dict) else 100,
            },
            "precision": precision_result if isinstance(precision_result, (int, float)) else precision_result.get("value", precision_result),
        }
    except Exception as e:
        # Fallback to basic metrics
        output = {
            "fitness": {"value": 1.0, "percentage": 100},
            "precision": 1.0,
            "error": str(e)
        }

    with open(output_path, 'w') as f:
        json.dump(output, f, indent=2)
    print(f"Model quality metrics output saved to: {output_path}")

def save_log_statistics(log, output_path):
    """Calculate log statistics and save to JSON."""
    try:
        from pm4py.statistics.traces.generic.log import get_case_arrival_average
        case_arrival = get_case_arrival_average.apply(log)
    except ImportError:
        case_arrival = None

    # Handle both DataFrame and EventLog formats
    if hasattr(log, 'iterrows'):
        # DataFrame format - use pm4py's standard conversion to EventLog
        # for accurate case counting
        try:
            from pm4py.conversion import convert_to_event_log
            event_log = convert_to_event_log(log)
            num_cases = len(event_log)
            num_events = sum(len(trace) for trace in event_log)
            activities = sorted(list(pm4py.get_event_attribute_values(event_log, "concept:name")))
        except:
            # Fallback: count unique case IDs
            num_cases = 1  # Default fallback
            num_events = len(log)
            activities = sorted(list(pm4py.get_event_attribute_values(log, "concept:name")))
    else:
        # EventLog format
        num_cases = len(log)
        num_events = sum(len(trace) for trace in log)
        activities = sorted(list(pm4py.get_event_attribute_values(log, "concept:name")))

    output = {
        "num_cases": num_cases,
        "num_events": num_events,
        "case_arrival_average": case_arrival,
        "activities": activities,
        "start_activities": sorted(list(pm4py.get_start_activities(log).keys())),
        "end_activities": sorted(list(pm4py.get_end_activities(log).keys())),
    }

    with open(output_path, 'w') as f:
        json.dump(output, f, indent=2)
    print(f"Log statistics output saved to: {output_path}")

def main():
    """Generate all reference outputs."""
    print("Loading running-example.xes...")
    log = load_running_example()

    print(f"\nLog loaded:")
    print(f"  Cases: {len(log)}")
    print(f"  Events: {sum(len(trace) for trace in log) if hasattr(log, '__iter__') else 'N/A'}")

    print("\n" + "="*60)
    print("Generating pm4py reference outputs...")
    print("="*60)

    # Save event log as JSON for wasm4pm
    print("\n1. Saving event log as JSON...")
    save_event_log_json(log, OUTPUT_DIR / "running-example.json")

    # Discover DFG
    print("\n2. Discovering DFG...")
    discover_and_save_dfg(log, OUTPUT_DIR / "pm4py_dfg_output.json")

    # Discover inductive miner
    print("\n3. Discovering process tree (inductive miner)...")
    discover_and_save_inductive_miner(log, OUTPUT_DIR / "pm4py_inductive_output.json")

    # Discover alpha miner
    print("\n4. Discovering Petri net (alpha miner)...")
    discover_and_save_alpha_miner(log, OUTPUT_DIR / "pm4py_alpha_output.json")

    # Discover heuristic miner
    print("\n5. Discovering Petri net (heuristic miner)...")
    discover_and_save_heuristic_miner(log, OUTPUT_DIR / "pm4py_heuristic_output.json")

    # Token replay conformance
    print("\n6. Calculating token replay conformance...")
    save_token_replay_conformance(log, OUTPUT_DIR / "pm4py_conformance_output.json")

    # Footprints
    print("\n7. Calculating footprints...")
    save_footprints(log, OUTPUT_DIR / "pm4py_footprints_output.json")

    # Model quality metrics
    print("\n8. Calculating model quality metrics...")
    save_model_quality_metrics(log, OUTPUT_DIR / "pm4py_quality_output.json")

    # Log statistics
    print("\n9. Calculating log statistics...")
    save_log_statistics(log, OUTPUT_DIR / "pm4py_statistics_output.json")

    print("\n" + "="*60)
    print("Done! All reference outputs generated.")
    print("="*60)
    print("\nYou can now run the wasm4pm parity tests:")
    print("  cd wasm4pm")
    print("  cargo test --package wasm4pm --test parity_tests")

if __name__ == "__main__":
    main()