sourcecode 0.34.0__tar.gz → 0.35.0__tar.gz

{sourcecode-0.34.0 → sourcecode-0.35.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sourcecode
-Version: 0.34.0
+Version: 0.35.0
 Summary: Deterministic codebase context for AI coding agents
 License:                                  Apache License
                                    Version 2.0, January 2004

{sourcecode-0.34.0 → sourcecode-0.35.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "sourcecode"
-version = "0.34.0"
+version = "0.35.0"
 description = "Deterministic codebase context for AI coding agents"
 readme = "README.md"
 requires-python = ">=3.9"

{sourcecode-0.34.0 → sourcecode-0.35.0}/src/sourcecode/__init__.py

@@ -1,3 +1,3 @@
 """sourcecode — Deterministic codebase context maps for AI coding agents."""
 
-__version__ = "0.34.0"
+__version__ = "0.35.0"

{sourcecode-0.34.0 → sourcecode-0.35.0}/src/sourcecode/ast_extractor.py

@@ -373,7 +373,8 @@ def _ts_types(root: Any, src: bytes) -> list[TypeDefinition]:
                 continue
             name = _text(name_n, src)
             fields: list[TypeField] = []
-            body_n = _find_child(node, "object_type")
+            # "interface_body" in tree-sitter-typescript >= 0.21; "object_type" in older builds
+            body_n = _find_child(node, "interface_body", "object_type")
             if body_n:
                 for prop in _walk(body_n):
                     if prop.type in ("property_signature", "method_signature"):
@@ -385,7 +386,7 @@ def _ts_types(root: Any, src: bytes) -> list[TypeDefinition]:
                             required = not any(c.type == "?" for c in prop.children)
                             fields.append(TypeField(name=prop_name, type=type_text, required=required))
             extends: list[str] = []
-            heritage_n = _find_child(node, "extends_type_clause", "class_heritage")
+            heritage_n = _find_child(node, "extends_type_clause", "extends_clause", "class_heritage")
             if heritage_n:
                 for ext_n in _walk(heritage_n):
                     if ext_n.type == "type_identifier":
@@ -429,6 +430,25 @@ def _ts_hooks(root: Any, src: bytes) -> list[str]:
     return sorted(used)
 
 
+def _merge_imports(imports: list[ImportRecord]) -> list[ImportRecord]:
+    """Merge multiple ImportRecords with the same source into one.
+
+    Tree-sitter correctly captures `import { A }` and `import type { B }` from
+    the same module as two separate statements.  Merging them produces a compact,
+    predictable contract where each source appears exactly once.
+    """
+    merged: dict[str, ImportRecord] = {}
+    for imp in imports:
+        if imp.source in merged:
+            existing = merged[imp.source]
+            combined_symbols = sorted(set(existing.symbols) | set(imp.symbols))
+            kind = existing.kind if existing.kind != "side_effect" else imp.kind
+            merged[imp.source] = ImportRecord(source=imp.source, symbols=combined_symbols, kind=kind)
+        else:
+            merged[imp.source] = imp
+    return list(merged.values())
+
+
 def _extract_ts_js_tree_sitter(path: str, source: str, lang_obj: Any, language: str) -> FileContract:
     try:
         parser = _get_parser(lang_obj)
@@ -436,7 +456,7 @@ def _extract_ts_js_tree_sitter(path: str, source: str, lang_obj: Any, language:
         tree = parser.parse(src_bytes)
         root = tree.root_node
 
-        imports = _ts_imports(root, src_bytes)
+        imports = _merge_imports(_ts_imports(root, src_bytes))
         exports = _ts_exports(root, src_bytes)
         exported_names = {e.name for e in exports}
         functions = _ts_functions(root, src_bytes, exported_names)

{sourcecode-0.34.0 → sourcecode-0.35.0}/src/sourcecode/contract_pipeline.py

@@ -25,6 +25,23 @@ from sourcecode.schema import EntryPoint, MonorepoPackageInfo
 _MAX_FILES = 500      # hard cap on files extracted per run
 _SRC_EXTENSIONS: frozenset[str] = frozenset(_LANGUAGE_MAP.keys())
 
+# Role-based score adjustments applied after contract extraction.
+# Runtime roles get a boost; config/util are neutral or penalized.
+_ROLE_SCORE: dict[str, float] = {
+    "entrypoint": 0.15,
+    "service":    0.10,
+    "route":      0.10,
+    "api":        0.08,
+    "middleware": 0.06,
+    "store":      0.05,
+    "model":      0.05,
+    "hook":       0.05,
+    "component":  0.03,
+    "util":       0.00,
+    "config":    -0.10,
+    "unknown":    0.00,
+}
+
 RankStrategy = Literal["relevance", "centrality", "git-churn"]
 
 
@@ -206,9 +223,9 @@ class ContractPipeline:
         if changed_only:
             src_paths = [p for p in src_paths if p in changed_files]
 
-        # Apply max_files cap
-        if len(src_paths) > self.max_files:
-            # Pre-rank by static relevance to pick best candidates
+        # Apply max_files cap — bypass when symbol search to ensure defining files are found.
+        # A symbol query over a large repo needs all files; result set is small after filtering.
+        if symbol is None and len(src_paths) > self.max_files:
             src_paths = sorted(
                 src_paths,
                 key=lambda p: (p in entry_paths, scorer.score(p)),
@@ -255,23 +272,9 @@ class ContractPipeline:
         # 7. Rank
         contracts = self._rank(contracts, rank_by)
 
-        # 8. Symbol filter — keep files that export or use the symbol
+        # 8. Symbol filter — keep files that define or import the symbol
         if symbol:
-            symbol_contracts = [
-                c for c in contracts
-                if any(e.name == symbol for e in c.exports)
-                or any(f.name == symbol for f in c.functions)
-                or symbol in {t.name for t in c.types}
-            ]
-            # Also pull in direct importers (fan_in sourcing)
-            importer_paths = {
-                c.path for c in contracts
-                for imp in c.imports
-                if symbol in imp.symbols
-            }
-            importer_contracts = [c for c in contracts if c.path in importer_paths]
-            symbol_contracts = list({c.path: c for c in symbol_contracts + importer_contracts}.values())
-            contracts = sorted(symbol_contracts, key=lambda c: -c.relevance_score)
+            contracts = _filter_by_symbol(contracts, symbol)
 
         # 9. Entrypoints-only filter
         if entrypoints_only and not symbol:
@@ -323,6 +326,9 @@ class ContractPipeline:
         churn_score = min(churn.get(c.path, 0) / 20.0, 0.1)
         base += churn_score
 
+        # Role-based boost: runtime roles score higher than auxiliary
+        base += _ROLE_SCORE.get(c.role, 0.0)
+
         return min(1.0, base)
 
     def _rank(self, contracts: list[FileContract], rank_by: RankStrategy) -> list[FileContract]:
@@ -385,6 +391,48 @@ def _limit_symbols(contracts: list[FileContract], max_symbols: int) -> list[File
     return result
 
 
+# ---------------------------------------------------------------------------
+# Symbol-aware filter
+# ---------------------------------------------------------------------------
+
+def _filter_by_symbol(contracts: list[FileContract], symbol: str) -> list[FileContract]:
+    """Return contracts that define or import *symbol*.
+
+    Matching strategy:
+    1. Exact match on export/function/type names.
+    2. Case-insensitive fallback when exact match yields nothing.
+    3. Importer contracts: files that name the symbol in their imports.
+
+    Defining contracts are ranked first; importers follow.
+    """
+    def _defines(c: FileContract, sym: str, case: bool) -> bool:
+        cmp = (lambda a, b: a.lower() == b.lower()) if case else (lambda a, b: a == b)
+        return (
+            any(cmp(e.name, sym) for e in c.exports)
+            or any(cmp(f.name, sym) for f in c.functions)
+            or any(cmp(t.name, sym) for t in c.types)
+        )
+
+    def _imports(c: FileContract, sym: str, case: bool) -> bool:
+        if case:
+            sym_l = sym.lower()
+            return any(sym_l == s.lower() for imp in c.imports for s in imp.symbols)
+        return any(sym in imp.symbols for imp in c.imports)
+
+    # Exact match first
+    defining = [c for c in contracts if _defines(c, symbol, case=False)]
+    if not defining:
+        defining = [c for c in contracts if _defines(c, symbol, case=True)]
+
+    importer_paths = {c.path for c in contracts if _imports(c, symbol, case=len(defining) == 0)}
+    # Exclude files already in defining set
+    defining_paths = {c.path for c in defining}
+    importers = [c for c in contracts if c.path in importer_paths and c.path not in defining_paths]
+
+    merged = list({c.path: c for c in defining + importers}.values())
+    return sorted(merged, key=lambda c: (c.path not in defining_paths, -c.relevance_score))
+
+
 # ---------------------------------------------------------------------------
 # Dependency graph emission
 # ---------------------------------------------------------------------------

{sourcecode-0.34.0 → sourcecode-0.35.0}/src/sourcecode/relevance_scorer.py

@@ -104,6 +104,18 @@ _LOW_RUNTIME_STEMS: frozenset[str] = frozenset({
     "gruntfile", "gulpfile", "webpack.config", "vite.config",
     "rollup.config", "babel.config", "jest.config", "vitest.config",
     "tsconfig", "jsconfig", ".eslintrc", ".prettierrc", ".editorconfig",
+    # doc-site tooling
+    "rspress", "rspress.config", "docusaurus.config", "docusaurus",
+    "vuepress.config", "vuepress", "nextra.config",
+    "astro.config", "gatsby.config", "gatsby-config",
+    # build/workspace orchestration
+    "turbo", "turbo.config", "nx", "nx.config", "lerna",
+    "esbuild.config", "swc.config", "postcss.config",
+    "tailwind.config", "tailwind",
+    # storybook
+    "main.storybook", "preview.storybook",
+    # playwright / cypress / e2e
+    "playwright.config", "cypress.config",
 })
 
 _HIGH_VALUE_SUFFIXES: frozenset[str] = frozenset({
@@ -169,15 +181,16 @@ class RelevanceScorer:
         if (any(m in f"/{norm}/" for m in _TEST_DIR_MARKERS)
                 or any(fname.startswith(p.strip(".")) or p in fname
                        for p in _TEST_FILE_PATTERNS)):
-            base -= 0.25
+            base -= 0.30
 
-        # Config/tooling filename penalty
+        # Config/tooling filename penalty — stronger than before
         if stem.lower() in _LOW_RUNTIME_STEMS:
-            base -= 0.2
+            base -= 0.30
 
-        # Auxiliary dir penalty
+        # Auxiliary dir penalty (docs, examples, demos, fixtures, scripts…)
+        # Aggressive: these almost never belong in top-ranked agent context
         if self._is_auxiliary(norm):
-            base -= 0.2
+            base -= 0.40
 
         return max(0.0, min(1.0, base))
 

{sourcecode-0.34.0 → sourcecode-0.35.0}/src/sourcecode/serializer.py

@@ -964,7 +964,9 @@ def _contract_view_minimal(
     # Compact summary
     if sm.contract_summary is not None:
         cs = sm.contract_summary
-        degraded = bool(cs.method_breakdown.get("heuristic", 0))
+        # degraded only when tree-sitter is actually unavailable — not when individual
+        # files fall back due to parse errors or size limits.
+        degraded = any("tree_sitter_unavailable" in lim for lim in cs.limitations)
         summary: dict[str, Any] = {
             "files": cs.extracted_files,
             "total": cs.total_files,