x4d-devkit 0.45.0__tar.gz → 0.45.1__tar.gz

{x4d_devkit-0.45.0/x4d_devkit.egg-info → x4d_devkit-0.45.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: x4d-devkit
-Version: 0.45.0
+Version: 0.45.1
 Summary: Official X-4D external SDK and CLI for datasets, platform APIs, training bridges, and automation
 Author: windzu
 License-Expression: Apache-2.0

{x4d_devkit-0.45.0 → x4d_devkit-0.45.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "x4d-devkit"
-version = "0.45.0"
+version = "0.45.1"
 description = "Official X-4D external SDK and CLI for datasets, platform APIs, training bridges, and automation"
 readme = "README.md"
 license = "Apache-2.0"

{x4d_devkit-0.45.0 → x4d_devkit-0.45.1}/tests/test_sdk_api.py

@@ -11,7 +11,7 @@ from x4d_devkit.eval.matching import MetricData
 
 
 def test_package_version_is_current_release():
-    assert x4d_devkit.__version__ == "0.45.0"
+    assert x4d_devkit.__version__ == "0.45.1"
 
 
 def test_loader_exposes_and_enforces_only_current_schema(tmp_path):

x4d_devkit-0.45.1/tests/test_work_session_validation_reuse.py

@@ -0,0 +1,109 @@
+"""Identical metadata may reuse validation; changed or mutable inputs may not."""
+
+from collections import OrderedDict
+import copy
+import json
+
+import pytest
+
+from tests.test_work_session import _manifest
+import x4d_devkit.work_session as sessions
+import x4d_devkit.validation.validator as validator
+
+
+@pytest.fixture(autouse=True)
+def isolated_validation_cache(monkeypatch):
+    monkeypatch.setattr(sessions, "_validated_sessions", OrderedDict())
+    monkeypatch.setattr(sessions, "_validated_session_bytes", 0)
+
+
+@pytest.fixture
+def validations(monkeypatch):
+    calls = []
+    original = validator.validate_clip_payload
+
+    def counted(*args, **kwargs):
+        calls.append(None)
+        return original(*args, **kwargs)
+
+    monkeypatch.setattr(validator, "validate_clip_payload", counted)
+    return calls
+
+
+def test_repeated_open_and_identity_comparison_validate_identical_payload_once(validations):
+    payload = _manifest()
+    first = sessions.parse_clip_work_session(payload)
+    second = sessions.parse_clip_work_session(first.to_dict())
+    assert sessions.compare_work_session(first, second).same
+    assert len(validations) == 1
+    assert first is not second
+
+
+def test_returned_snapshots_and_caller_input_cannot_poison_validation():
+    payload = _manifest()
+    original = copy.deepcopy(payload)
+    first = sessions.parse_clip_work_session(payload)
+    first.meta["sensors"]["lidar"]["frame_id"] = "changed-by-caller"
+    first.samples[0]["timestamp"]["sec"] = 99
+    payload["calibrated_sensor"][0]["translation"]["x"] = 100.0
+    second = sessions.parse_clip_work_session(original)
+    assert second.to_dict() == original
+    second.annotations[0]["bbox_3d"]["size"]["width"] = 999.0
+    assert sessions.parse_clip_work_session(original).to_dict() == original
+
+
+@pytest.mark.parametrize("changed", ["annotation", "sample_data", "typed_manifest"])
+def test_changed_payload_is_revalidated_even_with_identical_revisions(changed):
+    payload = _manifest()
+    parsed = sessions.parse_clip_work_session(payload)
+    if changed == "annotation":
+        payload["annotations"][0]["bbox_3d"]["size"]["width"] = -1.0
+    elif changed == "sample_data":
+        payload["sample_data"][0]["sample_token"] = "missing-sample"
+    else:
+        parsed.meta["schema_version"] = "0.17"
+        payload = parsed
+    with pytest.raises(ValueError):
+        sessions.parse_clip_work_session(payload)
+
+
+def test_cache_does_not_coerce_tuple_tables_into_valid_json_lists():
+    payload = _manifest()
+    sessions.parse_clip_work_session(payload)
+    payload["samples"] = tuple(payload["samples"])
+    with pytest.raises(ValueError, match="samples must be a list"):
+        sessions.parse_clip_work_session(payload)
+
+
+def test_cache_does_not_coerce_integer_object_keys_into_strings():
+    payload = _manifest()
+    urls = {}
+    for index, row in enumerate(payload["sample_data"]):
+        urls[str(index)] = payload["asset_urls"][row["sample_data_token"]]
+        row["sample_data_token"] = str(index)
+    payload["asset_urls"] = urls
+    sessions.parse_clip_work_session(payload)
+    payload["asset_urls"] = {int(key): value for key, value in urls.items()}
+    with pytest.raises((ValueError, TypeError)):
+        sessions.parse_clip_work_session(payload)
+
+
+@pytest.mark.parametrize("budget", ["entries", "bytes"])
+def test_validation_reuse_has_bounded_lru_storage(monkeypatch, validations, budget):
+    first = _manifest()
+    second = copy.deepcopy(first)
+    second["business_metadata_revision"] = "sha256:other-business-metadata"
+    if budget == "entries":
+        monkeypatch.setattr(sessions, "_VALIDATED_SESSION_MAX_ENTRIES", 1)
+    else:
+        sizes = [len(json.dumps(value, sort_keys=True, ensure_ascii=False,
+                               separators=(",", ":")).encode())
+                 for value in (first, second)]
+        monkeypatch.setattr(sessions, "_VALIDATED_SESSION_MAX_BYTES", sum(sizes) - 1)
+    sessions.parse_clip_work_session(first)
+    sessions.parse_clip_work_session(second)
+    sessions.parse_clip_work_session(second)
+    assert len(validations) == 2
+    sessions.parse_clip_work_session(first)
+    assert len(validations) == 3
+    assert len(sessions._validated_sessions) == 1

{x4d_devkit-0.45.0 → x4d_devkit-0.45.1}/x4d_devkit/work_session.py

@@ -7,9 +7,12 @@ without requiring a prebuilt archive.
 from __future__ import annotations
 
 import copy
+from collections import OrderedDict
 from dataclasses import dataclass
+import json
 from pathlib import PurePosixPath
 import re
+import threading
 from typing import Any, Mapping
 from urllib.parse import urljoin
 
@@ -290,6 +293,82 @@ class WorkSessionCompareResult:
         return {"status": self.status, "changed": list(self.changed)}
 
 
+_VALIDATED_SESSION_MAX_ENTRIES = 16
+_VALIDATED_SESSION_MAX_BYTES = 32 * 1024 * 1024
+_validated_sessions: OrderedDict[bytes, ClipWorkSessionManifest] = OrderedDict()
+_validated_session_bytes = 0
+_validated_session_guard = threading.Lock()
+
+
+def _validation_payload(data: dict[str, Any]) -> bytes | None:
+    # JSON serialization must never coerce an invalid key or container into a
+    # previously validated payload. Other Python values take the normal parser.
+    pending: list[Any] = [data]
+    seen: set[int] = set()
+    while pending:
+        value = pending.pop()
+        if type(value) in (dict, list):
+            if id(value) in seen:
+                continue
+            seen.add(id(value))
+            if type(value) is dict:
+                if any(type(key) is not str for key in value):
+                    return None
+                pending.extend(value.values())
+            else:
+                pending.extend(value)
+        elif type(value) not in (str, int, float, bool, type(None)):
+            return None
+    try:
+        payload = json.dumps(
+            data, sort_keys=True, ensure_ascii=False, allow_nan=False,
+            separators=(",", ":"),
+        ).encode("utf-8")
+    except (ValueError, TypeError, UnicodeError, RecursionError):
+        return None
+    return payload if len(payload) <= _VALIDATED_SESSION_MAX_BYTES else None
+
+
+def _session_snapshot(session: ClipWorkSessionManifest) -> ClipWorkSessionManifest:
+    # Schema internals contain immutable mapping proxies; reconstruct their
+    # small public snapshot and isolate every mutable table from the cache.
+    schema = parse_clip_label_schema_snapshot(session.label_schema.to_dict())
+    return copy.deepcopy(session, {id(session.label_schema): schema})
+
+
+def _cached_session(data: dict[str, Any]) -> ClipWorkSessionManifest | None:
+    payload = _validation_payload(data)
+    if payload is None:
+        return None
+    with _validated_session_guard:
+        session = _validated_sessions.get(payload)
+        if session is not None:
+            _validated_sessions.move_to_end(payload)
+    return _session_snapshot(session) if session is not None else None
+
+
+def _remember_session(session: ClipWorkSessionManifest) -> None:
+    global _validated_session_bytes
+    # Bind the validated result, never a caller-owned dictionary that could
+    # have changed during parsing. Bytes equality covers the complete payload.
+    payload = _validation_payload(session.to_dict())
+    if payload is None:
+        return
+    snapshot = _session_snapshot(session)
+    with _validated_session_guard:
+        if payload in _validated_sessions:
+            _validated_sessions.move_to_end(payload)
+            return
+        _validated_sessions[payload] = snapshot
+        _validated_session_bytes += len(payload)
+        while (
+            len(_validated_sessions) > _VALIDATED_SESSION_MAX_ENTRIES
+            or _validated_session_bytes > _VALIDATED_SESSION_MAX_BYTES
+        ):
+            evicted, _ = _validated_sessions.popitem(last=False)
+            _validated_session_bytes -= len(evicted)
+
+
 def parse_clip_work_session(
     value: ClipWorkSessionManifest | Mapping[str, Any],
 ) -> ClipWorkSessionManifest:
@@ -301,6 +380,9 @@ def parse_clip_work_session(
         data = dict(value)
     else:
         raise TypeError("work-session payload must be an object")
+    cached = _cached_session(data)
+    if cached is not None:
+        return cached
     if data.get("contract_version") != WORK_SESSION_CONTRACT_VERSION:
         raise ValueError(
             "work-session contract_version must be "
@@ -562,7 +644,7 @@ def parse_clip_work_session(
     )
     if not report.ok:
         raise ValueError(f"work-session payload violates the current clip contract:\n{report}")
-    return ClipWorkSessionManifest(
+    session = ClipWorkSessionManifest(
         project_id=project_id,
         clip_id=clip_id,
         ready=ready,
@@ -587,6 +669,8 @@ def parse_clip_work_session(
         asset_urls=asset_urls,
         semantic_scene_asset_urls=semantic_scene_asset_urls,
     )
+    _remember_session(session)
+    return session
 
 
 def compare_work_session(

{x4d_devkit-0.45.0 → x4d_devkit-0.45.1/x4d_devkit.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: x4d-devkit
-Version: 0.45.0
+Version: 0.45.1
 Summary: Official X-4D external SDK and CLI for datasets, platform APIs, training bridges, and automation
 Author: windzu
 License-Expression: Apache-2.0

{x4d_devkit-0.45.0 → x4d_devkit-0.45.1}/x4d_devkit.egg-info/SOURCES.txt

@@ -55,6 +55,7 @@ tests/test_vector_map.py
 tests/test_vector_map_geometry.py
 tests/test_vector_map_session.py
 tests/test_work_session.py
+tests/test_work_session_validation_reuse.py
 x4d_devkit/__init__.py
 x4d_devkit/annotation_contract.py
 x4d_devkit/annotation_reference.py