sourcecode 5.3.0__tar.gz → 5.3.1__tar.gz

{sourcecode-5.3.0 → sourcecode-5.3.1}/CHANGELOG.md

@@ -2,6 +2,58 @@
 
 ## [Unreleased]
 
+## [5.3.1] — 2026-08-11
+
+**Six fixes from one field battery.** 5.3.0 was run over the 14 repositories in
+the field corpus — 22 commands each, plus the symbol-level commands and the
+edge-input battery. No command crashed on its own account: every non-zero exit
+was a well-formed input error. What the battery found instead is the class this
+series keeps returning to — two surfaces answering one question with two
+numbers, and a count published under the name of a population it does not count.
+
+### Fixed
+- **One admission authority for the route population.** `extract_java_endpoints`
+  filtered its Java files with `is_test_path` alone, while `find_java_files` —
+  which feeds the IR and therefore `spring-audit`, `impact`, `validation` and
+  `data-exposure` — also drops the `src/main` of a module that is itself test
+  infrastructure. Measured on jobrunr: `endpoints` published 3 annotation routes
+  from `quarkus-jobrunr-tests` that `spring-audit` could not see at all, 3
+  against `endpoints_analyzed: 0`, under the payload's own
+  `semantic_model_endpoints_note` declaring the two counts to be one figure. The
+  rule is now `repository_ir.is_test_source_file`, called by both builders
+  (jobrunr 14/3 → 12/0; every other repository in the corpus unchanged).
+- **A call site is an HTTP call only where an HTTP client is.** The Java
+  consumer-join matcher admitted `<receiver>.put|patch|delete|exchange("…")` on
+  any receiver — HTTP verbs, and three of the most common method names in Java.
+  Measured on eureka joined against open-banking-gateway: all 20 rows of
+  `calls_unresolved` were `Map.put` literals in one file, and a
+  `map.put("/x", v)` would have been published as a consumer calling `/x` — the
+  direction the module declares decisive. A file must now import or declare an
+  HTTP client type before its call sites are read.
+- **A snapshot that measured nothing reports unknown, not zero.**
+  `update_ris_api_surface` / `update_ris_spring_audit` write a stub snapshot so
+  their own section survives when no full one exists; `cold-start` read that stub
+  as a measurement. On alfresco-community-repo — 7 000+ Java files, 901
+  WebScripts — it answered `validation: {spring_detected: false,
+  controllers_found: 0, status: "valid"}` with every section empty. Now
+  `status: cold_start_incomplete`, `validation.status: empty_snapshot`, every
+  count `null`, with a statement naming what is unknown and how to build the
+  snapshot. A repository measured to have no Spring surface still reports its
+  zeroes — that is a fact.
+- **The scale signal counts files, and says files.** The headline sentence of
+  every Java summary ended with "N Java classes" while counting `.java` paths.
+  BroadleafCommerce is 2 659 Java files and 3 032 declared types, so the label
+  contradicted `repo-ir.analysis_meta.classes` in the same run.
+- **The human audit bundle prints counts, not Python objects.** The two
+  breakdown lines of the markdown report interpolated their dicts directly, so
+  the signed evidence bundle published `- Endpoint access: {'no_rule_matched':
+  676}` on keycloak, and a literal `None` for a section that was never measured.
+- **Bookkeeping never takes down a finished analysis.** `RunRecord.write` had no
+  guard on its terminal write, so a run directory that disappeared during the
+  analysis — a concurrent `ask cache clear`, a tmp reaper, an evicted cache —
+  raised `FileNotFoundError` out of the analysis scope's exit. Observed on
+  `audit-report`, which died with a traceback having already composed the report.
+
 ## [5.3.0] — 2026-08-11
 
 **The capability queue, built on the stable point.** 5.2.0 was the highest-scored

{sourcecode-5.3.0 → sourcecode-5.3.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sourcecode
-Version: 5.3.0
+Version: 5.3.1
 Summary: Persistent structural context and ultra-fast repeated analysis for AI coding agents
 License-File: LICENSE
 Keywords: agents,ai,codebase,context,developer-tools,llm
@@ -42,7 +42,7 @@ Description-Content-Type: text/markdown
 
 **Context · Impact · Migration · Architecture · Review — everything from one structural model.**
 
-![Version](https://img.shields.io/badge/version-5.3.0-blue)
+![Version](https://img.shields.io/badge/version-5.3.1-blue)
 ![Python](https://img.shields.io/badge/python-3.9%2B-green)
 
 > **ASK Engine** is the product. The CLI command is **`ask`**. The legacy **`sourcecode`**
@@ -126,7 +126,7 @@ brew tap haroundominique/sourcecode && brew install sourcecode
 # pip / pipx
 pipx install sourcecode        # or: pip install sourcecode
 
-ask version                    # ask 5.3.0 — and, on a build that has aged,
+ask version                    # ask 5.3.1 — and, on a build that has aged,
                                # how many releases have probably shipped since
 ```
 

{sourcecode-5.3.0 → sourcecode-5.3.1}/README.md

@@ -4,7 +4,7 @@
 
 **Context · Impact · Migration · Architecture · Review — everything from one structural model.**
 
-![Version](https://img.shields.io/badge/version-5.3.0-blue)
+![Version](https://img.shields.io/badge/version-5.3.1-blue)
 ![Python](https://img.shields.io/badge/python-3.9%2B-green)
 
 > **ASK Engine** is the product. The CLI command is **`ask`**. The legacy **`sourcecode`**
@@ -88,7 +88,7 @@ brew tap haroundominique/sourcecode && brew install sourcecode
 # pip / pipx
 pipx install sourcecode        # or: pip install sourcecode
 
-ask version                    # ask 5.3.0 — and, on a build that has aged,
+ask version                    # ask 5.3.1 — and, on a build that has aged,
                                # how many releases have probably shipped since
 ```
 

{sourcecode-5.3.0 → sourcecode-5.3.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "sourcecode"
-version = "5.3.0"
+version = "5.3.1"
 description = "Persistent structural context and ultra-fast repeated analysis for AI coding agents"
 readme = "README.md"
 requires-python = ">=3.9"

{sourcecode-5.3.0 → sourcecode-5.3.1}/src/sourcecode/__init__.py

@@ -4,4 +4,4 @@ ASK Engine is the product. ``ask`` is the canonical CLI command; ``sourcecode``
 the legacy compatibility alias and the Python/PyPI package name. See
 docs/PRODUCT_IDENTITY.md (normative)."""
 
-__version__ = "5.3.0"
+__version__ = "5.3.1"

{sourcecode-5.3.0 → sourcecode-5.3.1}/src/sourcecode/audit_report.py

@@ -147,6 +147,23 @@ def build_audit_report(
     return normalized
 
 
+def _render_breakdown(value: Any) -> str:
+    """A count-per-class map, written the way a reader of a report reads one.
+
+    The two breakdown lines were interpolated as Python objects, so the signed
+    human bundle published `{'no_rule_matched': 676}` on keycloak — and a literal
+    `None` for a section that was not measured at all, which is the one thing a
+    report of record must never say when it means "not measured".
+    """
+    if value is None:
+        return "not measured"
+    if isinstance(value, dict):
+        if not value:
+            return "none reported"
+        return ", ".join(f"`{k}` {v}" for k, v in sorted(value.items()))
+    return f"`{value}`"
+
+
 def render_markdown(report: dict[str, Any]) -> str:
     """Human report renderer over the JSON contract."""
     summary = report.get("summary") or {}
@@ -160,8 +177,8 @@ def render_markdown(report: dict[str, Any]) -> str:
         f"- Profile set: `{profile}`",
         f"- Total defects: `{read_count(summary, 'total_defects')}`"
         + ("  **(floor — this run was cut short)**" if summary.get("partial") else ""),
-        f"- Risk bands: `{summary.get('risk_bands')}`",
-        f"- Endpoint access: `{summary.get('endpoint_access')}`",
+        f"- Risk bands: {_render_breakdown(summary.get('risk_bands'))}",
+        f"- Endpoint access: {_render_breakdown(summary.get('endpoint_access'))}",
         "",
         "## Top Risks",
         "",

{sourcecode-5.3.0 → sourcecode-5.3.1}/src/sourcecode/cli.py

@@ -13240,7 +13240,10 @@ def cold_start_cmd(
     """Output Repository Intelligence Snapshot bootstrap context as JSON.
 
     Returns instantly from persisted RIS — zero re-analysis cost.
-    status: cold_start_ready | cold_start_stale | no_ris
+    status: cold_start_ready | cold_start_stale | cold_start_incomplete | no_ris
+    `cold_start_incomplete` means the snapshot is missing sections (or holds no
+    measurement at all — `validation.status: empty_snapshot`, every count
+    `null`): unknown, never zero.
 
     \b
     Note: Full output is large (~100K–200K tokens for medium repos).

{sourcecode-5.3.0 → sourcecode-5.3.1}/src/sourcecode/consumer_join.py

@@ -71,6 +71,38 @@ _METHOD_ATTR = re.compile(r"RequestMethod\.(GET|POST|PUT|PATCH|DELETE|HEAD|OPTIO
 #: annotation-shaped path on a server controller is a route, not a call.
 _DECLARATIVE_CLIENT = re.compile(r"@(\w*FeignClient|HttpExchange|RegisterRestClient)\b")
 
+#: `put`, `patch` and `delete` are HTTP verbs and they are also three of the most
+#: common method names in Java: `map.put("us-east-1", "us-east-1a")` matches the
+#: call shape exactly. Measured on eureka, joined against open-banking-gateway:
+#: all 20 "unresolved client calls" were `Map.put` literals in
+#: `AbstractAzToRegionMapper`, and a `map.put("/x", v)` would have been published
+#: as a real consumer call to `/x` — the direction this module declares decisive.
+#: So the shape alone does not admit a call site: the FILE must carry evidence
+#: that an HTTP client is in it. These are the published type names of the HTTP
+#: client libraries themselves, matched as a declaration or an import — evidence,
+#: never a rule about the calling code's own naming.
+_HTTP_CLIENT_TYPE = re.compile(
+    r"\b(?:Async|Test|OAuth2)?RestTemplate\b"
+    r"|\bWeb(?:Test)?Client\b"
+    r"|\bRestClient\b"
+    r"|\bOkHttpClient\b"
+    r"|\bHttpClient\b"
+    r"|\bClientHttpRequest\w*\b"
+    r"|\bResteasyClient\b"
+    r"|\bjakarta\.ws\.rs\.client\b|\bjavax\.ws\.rs\.client\b"
+    r"|\bWebResource\b|\bWebTarget\b"
+    r"|\bRequestSpecification\b"  # rest-assured
+)
+
+
+def _file_declares_http_client(text: str) -> bool:
+    """True when this file imports or declares a type that speaks HTTP as a client.
+
+    A file with no such type has no `restTemplate` to call, so a `.put("…")` in it
+    is some other API with the same method name.
+    """
+    return bool(_HTTP_CLIENT_TYPE.search(text) or _DECLARATIVE_CLIENT.search(text))
+
 _JAVA_SKIP_DIRS = frozenset({
     "target", "build", "out", ".git", "node_modules", ".gradle",
 })
@@ -103,7 +135,10 @@ def extract_java_client_calls(base: Path) -> "tuple[list[ClientCall], list[dict]
         except OSError:
             continue
         rel = path.relative_to(base).as_posix()
-        for match in _JAVA_TEMPLATE_CALL.finditer(text):
+        _http_client_file = _file_declares_http_client(text)
+        for match in (
+            _JAVA_TEMPLATE_CALL.finditer(text) if _http_client_file else ()
+        ):
             op = match.group("op").lower()
             verb = next((v for k, v in _VERB_FROM_OP.items() if op.startswith(k)), "")
             if not verb and op == "exchange":

{sourcecode-5.3.0 → sourcecode-5.3.1}/src/sourcecode/repository_ir.py

@@ -6173,6 +6173,56 @@ def _recover_openapi_spec_routes(
 _SEMANTIC_MODEL_ENDPOINT_SOURCES: frozenset[str] = frozenset({"annotation", "openapi-spec"})
 
 
+def is_test_source_file(rel_path: str) -> bool:
+    """Return True when *rel_path* is a test source and must not enter any surface.
+
+    One authority for the question both Java file-list builders ask —
+    `find_java_files` (which feeds the IR and therefore every CIR consumer:
+    spring-audit, impact, validation, data-exposure) and
+    `extract_java_endpoints` (which feeds the `endpoints` command). They asked it
+    separately until a field run on jobrunr made the divergence visible:
+    `endpoints` published 3 annotation routes from
+    `quarkus-jobrunr-tests/src/main/java/org/jobrunr/quarkus/it/` that
+    `spring-audit` could not see at all, while the payload's own
+    `semantic_model_endpoints_note` claims the two counts are the same figure —
+    3 against `endpoints_analyzed: 0`.
+
+    Two facts, and C1-28 is why they are written as two.
+    `is_test_path` answers "is this file a test source"; since the authority took
+    over it never fires inside a declared main source root, which is correct
+    everywhere except inside a module that is itself test infrastructure — a
+    `test-framework` module has no product for its `src/main` to belong to. There
+    the weaker question is the right one: does this file look like a test by name
+    or by package. Both come from the one authority, so this stays a composition
+    of published facts rather than a third rule.
+
+    A test *fixture* module (`*-test-utils`) is deliberately NOT excluded here:
+    its endpoints are kept as raw evidence and tagged `scope: test_util`
+    downstream, by both surfaces.
+    """
+    rel = rel_path.replace("\\", "/").lstrip("/")
+    if _is_test_path(rel):
+        return True
+    _rel_sl = "/" + rel  # leading slash: the check below is prefix-safe
+    if "/src/main/" in _rel_sl and _matches_test_naming(rel):
+        _prefix = _rel_sl.split("/src/main/")[0]
+        _prefix_parts = [p for p in _prefix.split("/") if p]
+        # A module is a test module if is_test_path says so OR if any module
+        # directory component starts with "test" (e.g. "test-framework",
+        # "test-providers", "testsuite") OR contains "test"/"tests" as a
+        # hyphen/underscore-separated word (e.g. "jobrunr-micronaut-tests").
+        return (
+            _is_test_path(_prefix + "/x.java")
+            or any(p.lower().startswith("test") for p in _prefix_parts)
+            or any(
+                w in {"test", "tests", "spec", "specs"}
+                for p in _prefix_parts
+                for w in p.lower().replace("-", " ").replace("_", " ").split()
+            )
+        )
+    return False
+
+
 # ---------------------------------------------------------------------------
 # Convenience: find Java files in a repo
 # ---------------------------------------------------------------------------
@@ -6190,7 +6240,7 @@ def extract_java_endpoints(root: Path) -> "dict[str, Any]":
     """
     import re as _re
     from typing import Any as _Any
-    from sourcecode.path_filters import is_test_path, is_test_fixture_module_path
+    from sourcecode.path_filters import is_test_fixture_module_path
 
     _EXTENDS_FROM_SIG = _re.compile(r'\bextends\s+(\w+)')
 
@@ -6210,9 +6260,20 @@ def extract_java_endpoints(root: Path) -> "dict[str, Any]":
     _CLIENT_PATH_FRAGMENTS = (
         "/admin-client/", "/rest-client/", "/client-api/", "/api-client/",
     )
+    # Admission is `is_test_source_file`, the same authority `find_java_files`
+    # uses, so this command's route population and the IR's cannot diverge:
+    # `is_test_path` alone let a test MODULE's `src/main` resources through here
+    # while the IR dropped them, and the two published counts then contradicted
+    # each other (jobrunr: 3 annotation routes vs `endpoints_analyzed: 0`).
+    def _rel(p: Path) -> str:
+        try:
+            return str(p.relative_to(root)).replace("\\", "/")
+        except ValueError:
+            return str(p).replace("\\", "/")
+
     java_files = sorted(
         p for p in root.rglob("*.java")
-        if not is_test_path(str(p).replace("\\", "/"))
+        if not is_test_source_file(_rel(p))
         and "target/" not in str(p).replace("\\", "/")
         and not any(f in str(p).replace("\\", "/") for f in _CLIENT_PATH_FRAGMENTS)
     )
@@ -6841,43 +6902,9 @@ def find_java_files(
         except ValueError:
             continue
         parts = rel.split("/")
-        # Skip test dirs — use centralised is_test_path (consistent with
-        # extract_java_endpoints), guarded against false positives where a Java
-        # *package* is named "test" inside a production src/main/ source root.
-        if not include_tests:
-            # Two facts, and C1-28 is why they are now written as two.
-            # `is_test_path` answers "is this file a test source"; since the
-            # authority took over it never fires inside a declared main source
-            # root, which is correct everywhere except inside a module that is
-            # itself test infrastructure — a `test-framework` module has no
-            # product for its `src/main` to belong to. There the weaker question
-            # is the right one: does this file look like a test by name or by
-            # package. Both come from the one authority, so this stays a
-            # composition of published facts rather than a third rule. A test
-            # *fixture* module (`*-test-utils`) is deliberately NOT excluded
-            # here: its endpoints are kept as raw evidence and tagged
-            # `scope: test_util` downstream.
-            _rel_sl = "/" + rel  # leading slash: the checks below are prefix-safe
-            if _is_test_path(rel):
-                continue
-            if "/src/main/" in _rel_sl and _matches_test_naming(rel):
-                _prefix = _rel_sl.split("/src/main/")[0]
-                _prefix_parts = [p for p in _prefix.split("/") if p]
-                # A module is a test module if is_test_path says so OR if any
-                # module directory component starts with "test" (e.g. "test-framework",
-                # "test-providers", "testsuite") OR contains "test"/"tests" as a
-                # hyphen/underscore-separated word (e.g. "jobrunr-micronaut-tests").
-                _prefix_is_test = (
-                    _is_test_path(_prefix + "/x.java")
-                    or any(p.lower().startswith("test") for p in _prefix_parts)
-                    or any(
-                        w in {"test", "tests", "spec", "specs"}
-                        for p in _prefix_parts
-                        for w in p.lower().replace("-", " ").replace("_", " ").split()
-                    )
-                )
-                if _prefix_is_test:
-                    continue
+        # Skip test dirs — one authority, shared with extract_java_endpoints.
+        if not include_tests and is_test_source_file(rel):
+            continue
         # Skip vendor/generated/build dirs
         if any(part in _VENDOR_DIRS for part in parts[:-1]):
             continue

{sourcecode-5.3.0 → sourcecode-5.3.1}/src/sourcecode/ris.py

@@ -484,6 +484,20 @@ def _has_uncommitted_changes(repo_root: Path) -> bool:
     return bool(worktree_dirty(root, ignore=own))
 
 
+def _compact_probe(section: object) -> bool:
+    """True when an RIS section holds something a reader could act on.
+
+    A stub writes `{"endpoints": [], "controllers_index": []}` — a truthy dict
+    that contains no measurement, so the section's emptiness has to be asked of
+    its values, not of the container.
+    """
+    if isinstance(section, dict):
+        return any(_compact_probe(v) for v in section.values())
+    if isinstance(section, (list, tuple, set)):
+        return bool(section)
+    return section not in (None, "", 0, False)
+
+
 def get_cold_start_context(repo_root: Path) -> dict:
     """Return a lightweight bootstrap object from the persisted RIS.
 
@@ -527,22 +541,54 @@ def get_cold_start_context(repo_root: Path) -> dict:
         _endpoints_found = len(endpoints)
         # Spring is detected when controllers exist in structural map or api surface.
         _spring_detected = bool(_controllers_found) or bool(_controllers_in_api)
-        _validation_status = (
-            "incomplete_snapshot"
-            if _is_java and _spring_detected and _endpoints_found == 0
-            else "valid"
+        # A stub snapshot measured nothing. `update_ris_api_surface` /
+        # `update_ris_spring_audit` write one (`partial: True`, `confidence: 0.0`,
+        # `git_head: ""`) so their own section survives when no full snapshot
+        # exists yet — and on a repository whose Spring surface is genuinely
+        # empty, that stub is all there is. Measured on
+        # alfresco-community-repo (7 000+ Java files, 901 WebScripts): every
+        # section empty, and `validation: {spring_detected: false,
+        # controllers_found: 0, status: "valid"}` — an answer a bootstrapping
+        # agent reads as "this repository has no controllers and no entry
+        # points". Nothing was measured, so nothing is being reported: the
+        # sections are unknown, not zero.
+        _nothing_measured = not (
+            _compact_probe(ris.compact_summary)
+            or _compact_probe(ris.structural_map)
+            or _controllers_found
+            or _endpoints_found
+            or _compact_probe(ris.git_context_snapshot)
         )
+        if _nothing_measured:
+            _validation_status = "empty_snapshot"
+        elif _is_java and _spring_detected and _endpoints_found == 0:
+            _validation_status = "incomplete_snapshot"
+        else:
+            _validation_status = "valid"
         _validation: dict = {
-            "spring_detected": _spring_detected,
-            "controllers_found": _controllers_found,
-            "endpoints_found": _endpoints_found,
+            "spring_detected": None if _nothing_measured else _spring_detected,
+            "controllers_found": None if _nothing_measured else _controllers_found,
+            "endpoints_found": None if _nothing_measured else _endpoints_found,
             "status": _validation_status,
         }
+        if _nothing_measured:
+            _validation["statement"] = (
+                "This snapshot holds no measurement: it was written by a command "
+                "that persisted its own section only. Every field above is unknown "
+                "— not zero, and not evidence that the repository has no controllers, "
+                "endpoints or entry points. Build one with `ask cache warm <path>` "
+                "(or any analysis command) and re-run."
+            )
 
         # When the snapshot is structurally incomplete, downgrade status so agents
         # don't assume cold_start_ready when critical sections are missing.
         _status_base = "cold_start_stale" if stale else "cold_start_ready"
-        if _validation_status == "incomplete_snapshot" and not stale:
+        if _validation_status in ("incomplete_snapshot", "empty_snapshot") and not stale:
+            _status_base = "cold_start_incomplete"
+        if _validation_status == "empty_snapshot":
+            # A stub carries no head, so the staleness comparison it loses is not
+            # the fact worth reporting: `cold_start_stale` would send a reader to
+            # re-run against a newer tree, when what is missing is the snapshot.
             _status_base = "cold_start_incomplete"
 
         _compact = ris.compact_summary or {}

{sourcecode-5.3.0 → sourcecode-5.3.1}/src/sourcecode/runs.py

@@ -119,9 +119,28 @@ class RunRecord:
             payload["output_size_bytes"] = output_size
         if error:
             payload["error"] = error
+        # Bookkeeping never takes down an answer. `RunRecord.create` already
+        # treats an unwritable directory as "no record"; the terminal write did
+        # not, so a run directory that disappeared *during* the analysis — a
+        # concurrent `ask cache clear`, a tmp reaper, an evicted cache — raised
+        # FileNotFoundError out of the `_expensive_analysis_scope` exit, after the
+        # analysis had completed. Observed: `ask audit-report` on langchain4j died
+        # with a traceback over a missing `runs/<id>.tmp`, having already built
+        # the report. The record is recovery metadata; the answer is the product.
+        blob = json.dumps(payload, indent=2, sort_keys=True)
         tmp = self.path.with_suffix(".tmp")
-        tmp.write_text(json.dumps(payload, indent=2, sort_keys=True), encoding="utf-8")
-        tmp.replace(self.path)
+        for attempt in (0, 1):
+            try:
+                tmp.write_text(blob, encoding="utf-8")
+                tmp.replace(self.path)
+                return
+            except OSError:
+                if attempt:
+                    return
+                try:
+                    self.path.parent.mkdir(parents=True, exist_ok=True)
+                except OSError:
+                    return
 
 
 def classify_run(path: Path) -> dict[str, Any]:

{sourcecode-5.3.0 → sourcecode-5.3.1}/src/sourcecode/summarizer.py

@@ -389,14 +389,20 @@ class ProjectSummarizer:
             parts.append(f"Domains: {', '.join(domains)}")
 
         # Quantitative structural suffix for Java projects — adds concrete scale signals
-        # that README descriptions omit (class count, transactional boundary count).
+        # that README descriptions omit (file count, transactional boundary count).
         if primary is not None and primary.stack.lower() == "java":
             quant_parts: list[str] = []
             java_files = sum(
                 1 for p in sm.file_paths if p.endswith(".java")
             )
             if java_files >= 50:
-                quant_parts.append(f"{java_files:,} Java classes")
+                # This counts files, and says so. It read "N Java classes" — the
+                # headline sentence of every summary, naming a population the tool
+                # measures separately and differently: BroadleafCommerce is 2 659
+                # Java files and 3 032 declared types (2 174 classes, 742
+                # interfaces), so the label was wrong by a nested type per file
+                # and contradicted `repo-ir.analysis_meta.classes`.
+                quant_parts.append(f"{java_files:,} Java files")
             txn_classes: list[str] = []
             for stack in non_tooling_stacks:
                 txn_classes.extend(getattr(stack, "transactional_classes", []))