analysis

def find_hotspots(self, top_k: int = 10) -> list[dict[str, Any]]:
    """Find code hotspots using betweenness centrality.

    Hotspots are code elements that lie on many shortest paths
    between other elements - they are often bottlenecks or
    central integration points.

    Args:
        top_k: Number of top hotspots to return

    Returns:
        List of dictionaries with node info and betweenness score
    """
    view = self.graph.get_view([EdgeType.CALLS, EdgeType.REFERENCES])

    if view.number_of_nodes() == 0:
        return []

    try:
        betweenness = nx.betweenness_centrality(view, weight="weight")
    except nx.NetworkXError:
        return []

    return self._format_ranked_results(betweenness, top_k)

def find_foundations(self, top_k: int = 10) -> list[dict[str, Any]]:
    """Find foundational code using PageRank.

    Foundations are code elements that are heavily depended upon
    by other important code - the core infrastructure.

    Args:
        top_k: Number of top foundations to return

    Returns:
        List of dictionaries with node info and PageRank score
    """
    view = self.graph.get_view([EdgeType.CALLS, EdgeType.IMPORTS])

    if view.number_of_nodes() == 0:
        return []

    try:
        pagerank = nx.pagerank(view, alpha=0.85, weight="weight")
    except nx.NetworkXError:
        return []

    return self._format_ranked_results(pagerank, top_k)

def find_trusted_foundations(
    self,
    seed_nodes: list[str] | None = None,
    top_k: int = 10,
) -> list[dict[str, Any]]:
    """Find foundational code using TrustRank (noise-resistant PageRank).

    TrustRank propagates trust from seed nodes, making it more resistant
    to noise than standard PageRank. If no seed nodes provided, uses
    entry points as seeds.

    Args:
        seed_nodes: List of trusted node IDs (defaults to entry points)
        top_k: Number of top results to return

    Returns:
        List of dictionaries with node info and trust score
    """
    view = self.graph.get_view([EdgeType.CALLS, EdgeType.IMPORTS])

    if view.number_of_nodes() == 0:
        return []

    # Use entry points as default seeds
    if not seed_nodes:
        entry_points = self.find_entry_points()
        seed_nodes = [ep["id"] for ep in entry_points[:5]]

    if not seed_nodes:
        return self.find_foundations(top_k)

    # Build personalization dict for TrustRank
    trust = dict.fromkeys(view.nodes(), 0.0)
    for seed in seed_nodes:
        if seed in trust:
            trust[seed] = 1.0 / len(seed_nodes)

    try:
        scores = nx.pagerank(view, alpha=0.85, personalization=trust, weight="weight")
    except nx.NetworkXError:
        return []

    return self._format_ranked_results(scores, top_k)

def find_entry_points(self) -> list[dict[str, Any]]:
    """Find likely entry points in the code.

    Entry points are nodes with no incoming call edges but
    outgoing calls - they initiate execution flow.

    Returns:
        List of dictionaries with entry point node info
    """
    view = self.graph.get_view([EdgeType.CALLS])

    entry_points = []
    for node in view.nodes():
        in_deg = view.in_degree(node)
        out_deg = view.out_degree(node)

        # Entry point: no callers but makes calls
        if in_deg == 0 and out_deg > 0:
            node_data = self.graph.get_node_data(node)
            entry_points.append(
                {
                    "id": node,
                    "out_degree": out_deg,
                    **(node_data or {}),
                },
            )

    # Also check for main/run/start patterns
    for node, data in self.graph.nodes(data=True):
        name = str(data.get("name", "")).lower()
        if any(p in name for p in ("main", "__main__", "run", "start", "app", "cli")):
            if not any(ep["id"] == node for ep in entry_points):
                entry_points.append(
                    {
                        "id": node,
                        "out_degree": view.out_degree(node) if view.has_node(node) else 0,
                        **data,
                    },
                )

    # Sort by out_degree (more calls = more significant entry point)
    entry_points.sort(key=lambda x: x.get("out_degree", 0), reverse=True)

    return entry_points

def detect_modules(self, resolution: float = 1.0) -> list[dict[str, Any]]:
    """Detect logical modules using Louvain community detection.

    Uses the Louvain algorithm to find communities of densely
    connected code elements.

    Args:
        resolution: Clustering resolution (< 1 = larger clusters, > 1 = smaller)

    Returns:
        List of module dictionaries with members and metrics
    """
    view = self.graph.get_view([EdgeType.CALLS, EdgeType.IMPORTS])

    if view.number_of_nodes() < 2:
        return []

    # Louvain requires undirected graph
    undirected = view.to_undirected()

    try:
        # Try Leiden first (better community quality, requires backend)
        communities = nx.community.leiden_communities(undirected, resolution=resolution, seed=42)
    except (NotImplementedError, nx.NetworkXError, ValueError, RuntimeError):
        try:
            # Fallback to Louvain (pure NetworkX)
            communities = nx.community.louvain_communities(undirected, resolution=resolution, seed=42)
        except (nx.NetworkXError, ValueError, RuntimeError):
            return []

    modules = []
    for i, community in enumerate(communities):
        community_list = list(community)

        # Get key nodes (highest PageRank within community)
        subgraph = view.subgraph(community_list)
        if subgraph.number_of_nodes() > 0:
            try:
                local_pr = nx.pagerank(subgraph)
                key_nodes = sorted(local_pr.items(), key=lambda x: x[1], reverse=True)[:3]
            except (nx.NetworkXError, ValueError, RuntimeError):
                key_nodes = [(n, 0) for n in community_list[:3]]
        else:
            key_nodes = []

        # Calculate cohesion (internal/external edge ratio)
        cohesion = self._calculate_cohesion(view, community)

        modules.append(
            {
                "module_id": i,
                "size": len(community_list),
                "key_nodes": [{"id": n, "score": s} for n, s in key_nodes],
                "members": community_list,
                "cohesion": cohesion,
            },
        )

    # Sort by size (largest modules first)
    modules.sort(key=lambda x: x["size"], reverse=True)

    return modules

def find_clusters_by_pattern(self, rule_id: str) -> list[dict[str, Any]]:
    """Find clusters of nodes matching a specific AST-grep rule.

    Groups nodes by their rule_id metadata to find related
    business logic patterns.

    Args:
        rule_id: The rule identifier to filter by

    Returns:
        List of matching nodes grouped by file
    """
    matching_nodes: dict[str, list[dict[str, Any]]] = {}

    for node_id, data in self.graph.nodes(data=True):
        if data.get("rule_id") == rule_id:
            file_path = data.get("file_path", "unknown")
            if file_path not in matching_nodes:
                matching_nodes[file_path] = []
            matching_nodes[file_path].append({"id": node_id, **data})

    return [{"file": f, "matches": m, "count": len(m)} for f, m in matching_nodes.items()]

def find_clusters_by_category(self, category: str) -> list[dict[str, Any]]:
    """Find all nodes matching a business logic category.

    Args:
        category: Category to filter by (e.g., "db", "auth", "http")

    Returns:
        List of matching nodes with their locations
    """
    matches = []

    for node_id, data in self.graph.nodes(data=True):
        if data.get("category") == category:
            matches.append({"id": node_id, **data})

    return matches

def find_triangles(self, top_k: int = 10) -> list[dict[str, Any]]:
    """Find tightly-coupled code triads using triangle detection.

    Triangles in the call/import graph indicate three pieces of code
    that all depend on each other — potential cohesion or coupling issues.

    Args:
        top_k: Maximum number of triangles to return

    Returns:
        List of triangle dictionaries with the three node IDs
    """
    view = self.graph.get_view([EdgeType.CALLS, EdgeType.IMPORTS])
    undirected = view.to_undirected()

    triangles = []
    try:
        for triangle in nx.enumerate_all_cliques(undirected):
            if len(triangle) == 3:
                triangles.append(
                    {
                        "nodes": list(triangle),
                        "node_details": [{"id": n, **(self.graph.get_node_data(n) or {})} for n in triangle],
                    },
                )
                if len(triangles) >= top_k:
                    break
    except nx.NetworkXError:
        pass  # graph structure doesn't support triangle detection (e.g. directed)

    return triangles

def get_similar_nodes(self, node_id: str, top_k: int = 5) -> list[dict[str, Any]]:
    """Find nodes similar to a given node based on graph structure.

    Uses personalized PageRank to find nodes closely related
    to the target node.

    Args:
        node_id: The node to find similar nodes for
        top_k: Number of similar nodes to return

    Returns:
        List of similar nodes with similarity scores
    """
    view = self.graph.get_view()

    if not view.has_node(node_id):
        return []

    try:
        # Personalized PageRank with target node as seed
        ppr = nx.pagerank(view, personalization={node_id: 1}, alpha=0.85)
    except nx.NetworkXError:
        return []

    # Remove self, sort by score
    del ppr[node_id]
    ranked = sorted(ppr.items(), key=lambda x: x[1], reverse=True)[:top_k]

    return [{"id": n, "similarity": s, **(self.graph.get_node_data(n) or {})} for n, s in ranked if s > 0]

def calculate_coupling(self, node_a: str, node_b: str) -> dict[str, Any]:
    """Calculate coupling strength between two nodes.

    Considers shared neighbors, direct edges, and path length.

    Args:
        node_a: First node ID
        node_b: Second node ID

    Returns:
        Dictionary with coupling metrics
    """
    view = self.graph.get_view()

    if not view.has_node(node_a) or not view.has_node(node_b):
        return {"error": "Node not found", "coupling": 0.0}

    # Direct edge count
    direct_edges = 0
    if view.has_edge(node_a, node_b):
        direct_edges += 1
    if view.has_edge(node_b, node_a):
        direct_edges += 1

    # Shared neighbors
    neighbors_a = set(view.successors(node_a)) | set(view.predecessors(node_a))
    neighbors_b = set(view.successors(node_b)) | set(view.predecessors(node_b))
    shared = neighbors_a & neighbors_b

    # Shortest path length
    try:
        path_length = nx.shortest_path_length(view.to_undirected(), node_a, node_b)
    except nx.NetworkXNoPath:
        path_length = float("inf")

    # Calculate coupling score (higher = more coupled)
    coupling = direct_edges * 2.0 + len(shared) * 0.5 + (1.0 / (path_length + 1))

    return {
        "node_a": node_a,
        "node_b": node_b,
        "direct_edges": direct_edges,
        "shared_neighbors": len(shared),
        "path_length": path_length if path_length != float("inf") else None,
        "coupling": coupling,
    }

def get_dependency_chain(self, node_id: str, direction: str = "outgoing", max_depth: int = 5) -> dict[str, Any]:
    """Get the dependency chain from/to a node.

    Args:
        node_id: Starting node
        direction: "outgoing" (what this depends on) or "incoming" (what depends on this)
        max_depth: Maximum depth to traverse

    Returns:
        Dictionary with nodes and edges in the chain
    """
    view = self.graph.get_view([EdgeType.CALLS, EdgeType.IMPORTS])

    if not view.has_node(node_id):
        return {"error": "Node not found"}

    if direction == "outgoing":
        nodes = dict(nx.single_source_shortest_path_length(view, node_id, cutoff=max_depth))
    else:
        # Incoming: traverse reverse graph
        reverse = view.reverse()
        nodes = dict(nx.single_source_shortest_path_length(reverse, node_id, cutoff=max_depth))

    # Get edges within the discovered nodes
    subgraph = view.subgraph(nodes.keys())
    edges = list(subgraph.edges(data=True))

    return {
        "root": node_id,
        "direction": direction,
        "depth": max_depth,
        "nodes": [{"id": n, "distance": d, **(self.graph.get_node_data(n) or {})} for n, d in nodes.items()],
        "edges": [{"source": u, "target": v, **d} for u, v, d in edges],
    }

def find_unused_symbols(
    self,
    node_types: list[str] | None = None,
    exclude_patterns: list[str] | None = None,
) -> list[dict[str, Any]]:
    """Find symbols with zero incoming cross-file references.

    Identifies functions, classes, and methods that are defined but
    never referenced from other files — dead code candidates.

    Args:
        node_types: Filter to specific types (default: function, class, method)
        exclude_patterns: Regex patterns to exclude from results

    Returns:
        List of unused symbol dicts with id, name, file_path, node_type
    """
    target_types = (
        set(node_types)
        if node_types
        else {
            NodeType.FUNCTION.value,
            NodeType.CLASS.value,
            NodeType.METHOD.value,
        }
    )
    default_excludes = [r"^test_", r"^_", r"__init__", r"__main__"]
    excludes = [re.compile(p) for p in (exclude_patterns or default_excludes)]

    view = self.graph.get_view([EdgeType.REFERENCES, EdgeType.CALLS, EdgeType.IMPORTS])

    unused = []
    for node_id, data in self.graph.nodes(data=True):
        if data.get("node_type") not in target_types:
            continue

        name = str(data.get("name", ""))
        if any(pat.search(name) for pat in excludes):
            continue

        node_file = data.get("file_path", "")
        if not node_file:
            continue

        # Count incoming edges from OTHER files
        cross_file_refs = 0
        if view.has_node(node_id):
            for pred in view.predecessors(node_id):
                pred_data = self.graph.get_node_data(pred)
                pred_file = (pred_data or {}).get("file_path", "")
                if pred_file and pred_file != node_file:
                    cross_file_refs += 1
                    break  # One is enough to disqualify

        if cross_file_refs == 0:
            unused.append(
                {
                    "id": node_id,
                    "name": name,
                    "file_path": node_file,
                    "node_type": data.get("node_type"),
                    "line_start": data.get("line_start", 0),
                },
            )

    unused.sort(key=lambda x: (x["file_path"], x.get("line_start", 0)))
    return unused

def find_refactoring_candidates(self, top_k: int = 10) -> list[dict[str, Any]]:  # noqa: C901
    """Identify refactoring opportunities by combining multiple signals.

    Combines:
    - Clone pairs (SIMILAR_TO edges) -> "extract shared helper"
    - Code smell pattern matches (rule_id contains "code_smell") -> structural issues
    - Unused symbols -> "dead code removal"

    Args:
        top_k: Maximum number of candidates to return

    Returns:
        Ranked list of refactoring candidates with type, files, and rationale.
    """
    candidates: list[dict[str, Any]] = []

    # 1. Clone groups from SIMILAR_TO edges
    similar_edges = self.graph.get_edges_by_type(EdgeType.SIMILAR_TO)
    clone_groups: dict[str, list[str]] = {}
    for source, target, data in similar_edges:
        key = f"{source}|{target}" if source < target else f"{target}|{source}"
        if key not in clone_groups:
            clone_groups[key] = [source, target]
            candidates.append(
                {
                    "type": "extract_helper",
                    "pattern": f"Duplicate code between {source} and {target}",
                    "files": [source, target],
                    "occurrence_count": 2,
                    "duplicated_lines": int(data.get("duplicated_lines", 0)),
                    "score": int(data.get("duplicated_lines", 5)) * 2.0,
                },
            )

    # 2. Code smell patterns
    smell_counts: dict[str, list[str]] = {}
    for node_id, data in self.graph.nodes(data=True):
        rule_id = data.get("rule_id", "")
        note = data.get("note", "")
        if "code_smell" in note or "code_smell" in rule_id:
            if rule_id not in smell_counts:
                smell_counts[rule_id] = []
            smell_counts[rule_id].append(data.get("file_path", node_id))

    for rule_id, files in smell_counts.items():
        candidates.append(
            {
                "type": "code_smell",
                "pattern": rule_id,
                "files": list(set(files)),
                "occurrence_count": len(files),
                "duplicated_lines": 0,
                "score": len(files) * 1.5,
            },
        )

    # 3. Unused symbols
    unused = self.find_unused_symbols()
    if unused:
        # Group by file
        by_file: dict[str, list[str]] = {}
        for sym in unused:
            fp = sym["file_path"]
            if fp not in by_file:
                by_file[fp] = []
            by_file[fp].append(sym["name"])

        for fp, names in by_file.items():
            candidates.append(
                {
                    "type": "dead_code",
                    "pattern": f"{len(names)} unused symbol(s) in {fp}",
                    "files": [fp],
                    "occurrence_count": len(names),
                    "duplicated_lines": 0,
                    "score": len(names) * 1.0,
                },
            )

    # Sort by score descending, return top_k
    candidates.sort(key=lambda x: x["score"], reverse=True)
    return candidates[:top_k]