graphitect 0.2.0__py3-none-any.whl

graphify/symbol_resolution.py

@@ -0,0 +1,556 @@
+"""Deterministic symbol indexing and conservative cross-file resolution helpers."""
+
+from __future__ import annotations
+
+import ast
+import re
+import unicodedata
+from dataclasses import dataclass
+from pathlib import Path
+from collections.abc import Sequence
+from typing import Any
+
+from graphify.ids import make_id as _shared_make_id
+from graphify.paths import disambiguate_ambiguous_candidates
+from graphify.security import sanitize_metadata
+
+
+
+@dataclass(frozen=True)
+class ImportedSymbol:
+    """A Python imported name that can be used as deterministic resolution evidence."""
+
+    local_name: str
+    imported_name: str
+    module_stem: str
+    source_file: str
+    source_location: str
+
+
+def normalise_callable_label(label: str) -> str:
+    """Normalize a node label into the key used for call resolution."""
+
+    return label.strip().strip("()").lstrip(".").lower()
+
+
+def node_is_resolvable_symbol(node: dict[str, Any]) -> bool:
+    """Return True when a node is suitable for deterministic symbol lookup.
+
+    Requires ``file_type == "code"`` as the positive gate — only code-class
+    nodes participate as call targets. ``_EXCLUDED_FILE_TYPES`` is kept as
+    defensive-in-depth against legacy data, but the primary guard is the
+    positive code check. Document/paper/image/concept nodes (e.g. a Markdown
+    heading whose label happens to match a code identifier) MUST NOT become
+    callees for a raw code call.
+    """
+
+    if node.get("file_type") != "code":
+        return False
+    label = str(node.get("label", "")).strip()
+    if not label:
+        return False
+    if label.endswith((".py", ".js", ".ts", ".tsx", ".java", ".go", ".rs")):
+        return False
+    return bool(normalise_callable_label(label))
+
+
+def build_label_index(nodes: list[dict[str, Any]]) -> dict[str, list[str]]:
+    """Build label -> node id list for conservative cross-file resolution."""
+
+    index: dict[str, list[str]] = {}
+    for node in nodes:
+        if not node_is_resolvable_symbol(node):
+            continue
+        node_id = node.get("id")
+        if not node_id:
+            continue
+        key = normalise_callable_label(str(node.get("label", "")))
+        if not key:
+            continue
+        index.setdefault(key, []).append(str(node_id))
+    return index
+
+
+def existing_edge_pairs(edges: list[dict[str, Any]]) -> set[tuple[str, str, str]]:
+    """Return all existing source/target/relation edge triples.
+
+    Includes relation so that a prior "contains" or "method" edge does not
+    suppress a semantically distinct "calls" edge between the same endpoints (#F5).
+    """
+
+    triples: set[tuple[str, str, str]] = set()
+    for edge in edges:
+        source = edge.get("source")
+        target = edge.get("target")
+        relation = edge.get("relation", "")
+        if source and target:
+            triples.add((str(source), str(target), str(relation)))
+    return triples
+
+
+def iter_raw_calls(per_file: Sequence[object]) -> list[dict[str, Any]]:
+    """Return raw calls from all per-file extraction fragments.
+
+    Parameter is ``Sequence[object]`` (not ``Sequence[dict[str, Any] | None]``)
+    because external extraction output may contain arbitrary deserialized
+    JSON. Defensive against malformed fragments: non-dict per-file entries
+    are skipped, non-list ``raw_calls`` are treated as empty, and non-dict
+    items inside the list are silently dropped. The downstream resolvers
+    assume every returned item is a dict and they expect this guarantee.
+    """
+
+    calls: list[dict[str, Any]] = []
+    for result in per_file:
+        if not isinstance(result, dict):
+            continue
+        raw_calls = result.get("raw_calls", [])
+        if not isinstance(raw_calls, list):
+            continue
+        for raw_call in raw_calls:
+            if isinstance(raw_call, dict):
+                calls.append(raw_call)
+    return calls
+
+
+def _module_stem(module_name: str | None) -> str:
+    """Return the final module component used to match Graphify source stems."""
+
+    if not module_name:
+        return ""
+    return module_name.strip(".").split(".")[-1]
+
+
+def parse_python_import_aliases(path: Path) -> dict[str, ImportedSymbol]:
+    """Parse deterministic Python import aliases from one source file.
+
+    Supported forms:
+        from helper import transform
+        from helper import transform as tx
+        from .helper import transform
+
+    The function deliberately does not resolve plain ``import helper`` member
+    calls because current raw call records do not preserve the receiver name from
+    ``helper.transform()``. That can be added later only after raw call facts are
+    extended to include the receiver expression.
+    """
+
+    try:
+        source = path.read_text(encoding="utf-8", errors="replace")
+        tree = ast.parse(source)
+    except (OSError, SyntaxError):
+        return {}
+
+    aliases: dict[str, ImportedSymbol] = {}
+    source_file = str(path)
+
+    # Only top-level `from ... import ...` statements count as file-wide
+    # evidence. Nested/function-local imports do NOT — they're only valid
+    # inside their lexical scope, and our raw-call records don't currently
+    # carry enough scope info to match the import site safely. Walking
+    # ast.walk(tree) would incorrectly justify calls in other scopes.
+    for node in tree.body:
+        if not isinstance(node, ast.ImportFrom):
+            continue
+        module_stem = _module_stem(node.module)
+        if not module_stem:
+            continue
+        for alias in node.names:
+            if alias.name == "*":
+                continue
+            local_name = alias.asname or alias.name
+            aliases[local_name] = ImportedSymbol(
+                local_name=local_name,
+                imported_name=alias.name,
+                module_stem=module_stem,
+                source_file=source_file,
+                source_location=f"L{getattr(node, 'lineno', 1)}",
+            )
+
+    return aliases
+
+
+def _node_source_stem(node: dict[str, Any]) -> str:
+    """Return the stem of a node's source file."""
+
+    source_file = str(node.get("source_file", ""))
+    if not source_file:
+        return ""
+    return Path(source_file).stem
+
+
+def build_python_symbol_index(nodes: list[dict[str, Any]]) -> dict[tuple[str, str], list[str]]:
+    """Build ``(module_stem, normalized_symbol_name) -> node_ids``.
+
+    This index is stricter than the global label index. It uses both the module
+    stem and the symbol label, which allows import evidence to resolve calls that
+    global label uniqueness alone cannot safely resolve.
+    """
+
+    index: dict[tuple[str, str], list[str]] = {}
+    for node in nodes:
+        if not node_is_resolvable_symbol(node):
+            continue
+        source_stem = _node_source_stem(node)
+        if not source_stem:
+            continue
+        label = normalise_callable_label(str(node.get("label", "")))
+        if not label:
+            continue
+        node_id = node.get("id")
+        if not node_id:
+            continue
+        index.setdefault((source_stem, label), []).append(str(node_id))
+    return index
+
+
+def find_unique_python_symbol(
+    symbol_index: dict[tuple[str, str], list[str]],
+    imported: ImportedSymbol,
+) -> str | None:
+    """Resolve one imported symbol to exactly one Graphify node id."""
+
+    candidates = symbol_index.get((imported.module_stem, imported.imported_name.lower()), [])
+    if len(candidates) == 1:
+        return candidates[0]
+    return None
+
+
+def resolve_python_import_guided_calls(
+    per_file: Sequence[object],
+    paths: Sequence[Path],
+    all_nodes: list[dict[str, Any]],
+    all_edges: list[dict[str, Any]],
+) -> list[dict[str, Any]]:
+    """Resolve raw Python calls using explicit import evidence.
+
+    Only ``from module import symbol [as alias]`` forms are handled. Member calls
+    remain skipped because the current raw call fact does not carry receiver
+    information.
+
+    Parameter ``per_file`` is ``Sequence[object]`` because external extraction
+    output may contain arbitrary deserialized JSON. Non-dict slots are
+    treated as empty fragments, and indices past ``len(per_file)`` are also
+    treated as empty (paths longer than per_file is tolerated).
+    """
+
+    symbol_index = build_python_symbol_index(all_nodes)
+    known_pairs = existing_edge_pairs(all_edges)
+    # Build result_by_file defensively:
+    #   - skip indices past the end of per_file (paths shorter than per_file
+    #     also OK; the zip-like behavior is what callers expect)
+    #   - non-dict per_file slots fall back to the empty fragment so the
+    #     downstream `.get("raw_calls", [])` lookup never raises
+    result_by_file: dict[str, dict[str, Any]] = {}
+    for index, path in enumerate(paths):
+        if path.suffix != ".py":
+            continue
+        slot: Any = per_file[index] if index < len(per_file) else None
+        result_by_file[str(path)] = slot if isinstance(slot, dict) else {"nodes": [], "edges": []}
+    resolved_edges: list[dict[str, Any]] = []
+
+    for path in paths:
+        if path.suffix != ".py":
+            continue
+        source_file = str(path)
+        aliases = parse_python_import_aliases(path)
+        if not aliases:
+            continue
+        file_result = result_by_file.get(source_file, {"raw_calls": []})
+        raw_calls = file_result.get("raw_calls", [])
+        if not isinstance(raw_calls, list):
+            continue
+        for raw_call in raw_calls:
+            if not isinstance(raw_call, dict):
+                continue
+            if raw_call.get("is_member_call"):
+                continue
+            callee = str(raw_call.get("callee", "")).strip()
+            if not callee:
+                continue
+            imported = aliases.get(callee)
+            if imported is None:
+                continue
+            target = find_unique_python_symbol(symbol_index, imported)
+            if target is None:
+                continue
+            caller = str(raw_call.get("caller_nid", ""))
+            if not caller or caller == target:
+                continue
+            pair = (caller, target, "calls")
+            if pair in known_pairs:
+                continue
+            known_pairs.add(pair)
+            resolved_edges.append(
+                {
+                    "source": caller,
+                    "target": target,
+                    "relation": "calls",
+                    "context": "import_guided_call",
+                    "confidence": "EXTRACTED",
+                    "confidence_score": 1.0,
+                    "source_file": raw_call.get("source_file", source_file),
+                    "source_location": raw_call.get("source_location") or imported.source_location,
+                    "weight": 1.0,
+                    "metadata": sanitize_metadata({
+                        "resolver": "python_import_guided",
+                        "local_name": imported.local_name,
+                        "imported_name": imported.imported_name,
+                        "module_stem": imported.module_stem,
+                        "import_source_location": imported.source_location,
+                    }),
+                }
+            )
+
+    return resolved_edges
+
+
+def resolve_cross_file_raw_calls(
+    per_file: Sequence[dict[str, Any] | None],
+    all_nodes: list[dict[str, Any]],
+    all_edges: list[dict[str, Any]],
+) -> list[dict[str, Any]]:
+    """Resolve unqualified raw calls conservatively after all files are known.
+
+    This intentionally preserves Graphify's existing behavior:
+    - member calls are skipped;
+    - ambiguous labels are skipped;
+    - only a single unique candidate is emitted;
+    - emitted edges are INFERRED because the raw call alone is not import proof.
+    """
+
+    label_index = build_label_index(all_nodes)
+    known_pairs = existing_edge_pairs(all_edges)
+    # nid -> source_file, for the shared god-node tie-breakers (#1553) so a
+    # same-named test mock no longer erases a real cross-file call.
+    nid_to_source_file = {
+        str(n.get("id")): str(n.get("source_file", ""))
+        for n in all_nodes
+        if n.get("id")
+    }
+    resolved: list[dict[str, Any]] = []
+
+    for raw_call in iter_raw_calls(per_file):
+        callee = str(raw_call.get("callee", "")).strip()
+        if not callee:
+            continue
+        if raw_call.get("is_member_call"):
+            continue
+        candidates = label_index.get(callee.lower(), [])
+        if not candidates:
+            continue
+        if len(candidates) == 1:
+            target: str | None = candidates[0]
+        else:
+            # Ambiguous bare name. Apply the shared tie-breakers (non-test
+            # preference, then path proximity); resolve only if exactly one
+            # candidate survives, else preserve the god-node guard and skip.
+            target = disambiguate_ambiguous_candidates(
+                candidates,
+                {c: nid_to_source_file.get(c, "") for c in candidates},
+                str(raw_call.get("source_file", "")),
+            )
+            if target is None:
+                continue
+        caller = str(raw_call.get("caller_nid", ""))
+        if not caller:
+            continue
+        if target == caller:
+            continue
+        pair = (caller, target, "calls")
+        if pair in known_pairs:
+            continue
+        known_pairs.add(pair)
+        resolved.append(
+            {
+                "source": caller,
+                "target": target,
+                "relation": "calls",
+                "context": "call",
+                "confidence": "INFERRED",
+                # 0.85, not 0.8 — the extraction-spec rubric's INFERRED values
+                # are a discrete set and 0.8 is not one of them (#2813).
+                "confidence_score": 0.85,
+                "source_file": raw_call.get("source_file", ""),
+                "source_location": raw_call.get("source_location"),
+                "weight": 1.0,
+            }
+        )
+
+    return resolved
+
+
+def _bash_make_id(*parts: str) -> str:
+    """Bash symbol node ID via the single shared recipe (#1378).
+
+    Previously an inline copy to dodge an import cycle; ``graphify.ids`` is
+    dependency-free, so it can be imported directly.
+    """
+    return _shared_make_id(*parts)
+
+
+from graphify.extractors.base import _file_stem as _bash_file_stem  # canonical recipe (no import cycle: base imports only graphify.ids)
+
+
+def _file_node_id_for_path(path: Path, root: Path) -> str:
+    # Produce the canonical {parent_dir}_{stem} file-node ID that extract()'s
+    # id_remap generates (#1033), so bash `source` edges land on the real file
+    # node instead of an orphan. _bash_make_id / _bash_file_stem are exact copies
+    # of extract._make_id / extract._file_stem, so IDs match.
+    try:
+        rel = path.resolve().relative_to(root.resolve())
+    except ValueError:
+        return _bash_make_id(str(path))  # path outside root: hash absolute path as fallback
+    return _bash_make_id(_bash_file_stem(rel))
+
+
+def resolve_bash_source_edges(
+    per_file: Sequence[dict | None],
+    paths: Sequence[Path],
+    root: Path,
+    existing_edges: list[dict] | None = None,
+) -> list[dict]:
+    """Resolve Bash source/import edges and source-backed function calls.
+
+    Defensive against malformed extraction fragments: non-dict ``per_file``
+    entries, missing ``bash_sources``/``raw_calls`` keys, non-dict items in
+    those lists, and missing/empty ``id`` / ``target_path`` / ``caller_nid``
+    fields all yield silent skips rather than ``KeyError``.
+
+    ``bash_sources[].target_path`` contract (Graphify static-analysis policy):
+        - Absolute paths: resolved as-is.
+        - Relative paths: resolved against the *source file's* directory
+          (i.e. ``Path(path).parent / target_path``).
+          NOTE: this is a deterministic static-analysis policy chosen by
+          Graphify, NOT bash runtime semantics. At runtime, ``source ./X``
+          is resolved against the shell's current working directory. We
+          prefer source-file-relative because static analysis cannot know
+          the future CWD; resolving against the file being analyzed gives
+          deterministic, reproducible edges across runs.
+        - Inputs of type ``str`` and ``pathlib.Path`` are processed.
+          Anything else is silently skipped.
+    """
+    path_by_index = [Path(p).resolve() for p in paths]
+    file_nid_by_path = {p: _file_node_id_for_path(p, root) for p in path_by_index}  # resolved paths only
+
+    functions_by_file: dict[str, dict[str, str]] = {}
+    for result, path in zip(per_file, path_by_index):
+        if not isinstance(result, dict):
+            continue
+        file_nid = file_nid_by_path[path]
+        nodes = result.get("nodes", [])
+        if not isinstance(nodes, list):
+            continue
+        for node in nodes:
+            if not isinstance(node, dict):
+                continue
+            metadata = node.get("metadata", {})
+            if not isinstance(metadata, dict):
+                continue
+            if metadata.get("kind") != "bash_function":
+                continue
+            name = str(node.get("label", "")).removesuffix("()").strip()
+            node_id = node.get("id")
+            if not name or not node_id:
+                continue
+            functions_by_file.setdefault(file_nid, {})[name] = str(node_id)
+
+    sourced_files: dict[str, set[str]] = {}
+    resolved_edges: list[dict] = []
+    existing = existing_edge_pairs(existing_edges or [])
+
+    for result, path in zip(per_file, path_by_index):
+        if not isinstance(result, dict):
+            continue
+        src_file_nid = file_nid_by_path[path]
+        bash_sources = result.get("bash_sources", [])
+        if not isinstance(bash_sources, list):
+            continue
+        for source in bash_sources:
+            if not isinstance(source, dict):
+                continue
+            raw_target = source.get("target_path")
+            if not isinstance(raw_target, (str, Path)) or not str(raw_target).strip():
+                continue
+            # Relative paths resolve against the source file's directory —
+            # Graphify static-analysis policy (NOT bash runtime semantics;
+            # at runtime `source ./X` is CWD-relative, but static analysis
+            # can't know the future CWD, so we resolve relative to the
+            # file being analyzed for deterministic, reproducible edges).
+            candidate = Path(raw_target)
+            if not candidate.is_absolute():
+                candidate = path.parent / candidate
+            try:
+                target_path = candidate.resolve()
+            except (OSError, RuntimeError):
+                continue
+            target_file_nid = file_nid_by_path.get(target_path)
+            if target_file_nid is None:
+                continue
+            sourced_files.setdefault(src_file_nid, set()).add(target_file_nid)
+            key = (src_file_nid, target_file_nid, "imports_from")
+            if key in existing:
+                continue
+            existing.add(key)
+            resolved_edges.append(
+                {
+                    "source": src_file_nid,
+                    "target": target_file_nid,
+                    "relation": "imports_from",
+                    "context": "import",
+                    "confidence": "EXTRACTED",
+                    "confidence_score": 1.0,
+                    "source_file": source.get("source_file", str(path)),
+                    "source_location": source.get("source_location", ""),
+                    "weight": 1.0,
+                }
+            )
+
+    for result, path in zip(per_file, path_by_index):
+        if not isinstance(result, dict):
+            continue
+        caller_file_nid = file_nid_by_path[path]
+        imported_file_ids = sourced_files.get(caller_file_nid, set())
+        if not imported_file_ids:
+            continue
+        raw_calls = result.get("raw_calls", [])
+        if not isinstance(raw_calls, list):
+            continue
+        for raw_call in raw_calls:
+            if not isinstance(raw_call, dict):
+                continue
+            if raw_call.get("language") != "bash":
+                continue
+            callee = raw_call.get("callee")
+            caller_nid = raw_call.get("caller_nid")
+            # callee must be a non-empty string — anything else (list, dict,
+            # int, None, …) is silently skipped to avoid TypeError on the
+            # `in functions_by_file[...]` membership check below.
+            if not isinstance(callee, str) or not callee or not caller_nid:
+                continue
+            matches = [
+                functions_by_file[file_nid][callee]
+                for file_nid in imported_file_ids
+                if callee in functions_by_file.get(file_nid, {})
+            ]
+            if len(matches) != 1:
+                continue
+            target = matches[0]
+            key = (str(caller_nid), target, "calls")
+            if key in existing:
+                continue
+            existing.add(key)
+            resolved_edges.append(
+                {
+                    "source": str(caller_nid),
+                    "target": target,
+                    "relation": "calls",
+                    "context": "call",
+                    "confidence": "EXTRACTED",
+                    "confidence_score": 1.0,
+                    "source_file": raw_call.get("source_file", str(path)),
+                    "source_location": raw_call.get("source_location", ""),
+                    "weight": 1.0,
+                }
+            )
+
+    return resolved_edges