x4d-devkit 0.50.0__tar.gz → 0.51.0__tar.gz

{x4d_devkit-0.50.0 → x4d_devkit-0.51.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: x4d-devkit
-Version: 0.50.0
+Version: 0.51.0
 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-0.50.0 → x4d_devkit-0.51.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "x4d-devkit"
-version = "0.50.0"
+version = "0.51.0"
 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-0.51.0/tests/test_annotation_document_hash.py

@@ -0,0 +1,111 @@
+from __future__ import annotations
+
+import dataclasses
+import hashlib
+import random
+
+import pytest
+import rfc8785
+
+from tests.test_annotation_contract import _annotation
+from x4d_devkit.annotation_contract import (
+    ANNOTATION_CONTRACT_VERSION,
+    AnnotationContractError,
+    ClipAnnotationDocument,
+    build_clip_annotations,
+    parse_clip_annotations,
+)
+
+_EDGE_FLOATS = (0.0, -0.0, 1e21, 1e-7, 5e-324, 1.7976931348623157e308, 0.1 + 0.2, -123.456)
+_EDGE_TEXT = ("", "é", "雪", "\U0001f697", "a\"b\\c", "
", "\x7f")
+
+
+def _sha(payload) -> str:
+    return f"sha256:{hashlib.sha256(rfc8785.dumps(payload)).hexdigest()}"
+
+
+def _random_rows(rng: random.Random, count: int) -> list[dict]:
+    rows = []
+    for index in range(count):
+        row = _annotation(
+            f"ann-{rng.randrange(10**9):09d}-{index}",
+            f"sample-{rng.randrange(5)}",
+            instance=f"instance-{rng.randrange(4)}",
+        )
+        row["bbox_3d"]["translation"]["x"] = rng.choice(_EDGE_FLOATS)
+        row["bbox_3d"]["size"]["width"] = rng.random() * 10
+        row["attributes"] = {rng.choice(_EDGE_TEXT) or "k": rng.choice(_EDGE_TEXT)}
+        if rng.random() < 0.5:
+            row["velocity"] = {"vx": rng.choice(_EDGE_FLOATS), "vy": 0.0, "vz": -1.5}
+        rows.append(row)
+    for instance in {row["instance_token"] for row in rows}:
+        motion = rng.choice(("auto", "static", "dynamic"))
+        for row in rows:
+            if row["instance_token"] == instance:
+                row["motion"] = motion
+    return rows
+
+
+def test_single_pass_digests_match_full_rfc8785_canonicalization():
+    rng = random.Random(8785)
+    for count in (0, 1, 2, 17, 120):
+        document = build_clip_annotations("clip-a", "base_link", _random_rows(rng, count))
+        payload = document.to_dict()
+        revision_payload = {key: value for key, value in payload.items() if key != "annotation_revision"}
+        assert document.annotation_revision == _sha(revision_payload)
+        assert document.document_hash() == _sha(payload)
+
+        parsed = parse_clip_annotations(payload)
+        assert parsed.annotation_revision == document.annotation_revision
+        assert parsed.document_hash() == _sha(payload)
+
+
+def test_document_hash_is_derived_for_directly_constructed_and_replaced_documents():
+    document = build_clip_annotations("clip-a", "base_link", _random_rows(random.Random(1), 9))
+    direct = ClipAnnotationDocument(
+        clip_id=document.clip_id,
+        annotation_frame_id=document.annotation_frame_id,
+        annotation_revision=document.annotation_revision,
+        annotations=document.annotations,
+    )
+    assert direct.document_hash() == document.document_hash()
+
+    replaced = dataclasses.replace(document, annotations=document.annotations[:3])
+    assert replaced.document_hash() == _sha(replaced.to_dict())
+    assert replaced.document_hash() != document.document_hash()
+
+
+def test_document_hash_is_not_part_of_document_equality():
+    rows = _random_rows(random.Random(2), 5)
+    built = build_clip_annotations("clip-a", "base_link", rows)
+    parsed = parse_clip_annotations(built.to_dict())
+    plain = ClipAnnotationDocument(
+        clip_id=built.clip_id,
+        annotation_frame_id=built.annotation_frame_id,
+        annotation_revision=built.annotation_revision,
+        annotations=built.annotations,
+    )
+    assert built == parsed == plain
+    assert "_document_hash" not in {field.name for field in dataclasses.fields(built)}
+    assert built.to_dict()["contract_version"] == ANNOTATION_CONTRACT_VERSION
+
+
+def test_parse_reuses_only_documents_this_module_verified():
+    built = build_clip_annotations("clip-a", "base_link", _random_rows(random.Random(3), 4))
+    assert parse_clip_annotations(built) is built
+    assert parse_clip_annotations(parse_clip_annotations(built.to_dict())) is not built
+
+    forged = ClipAnnotationDocument(
+        clip_id=built.clip_id,
+        annotation_frame_id=built.annotation_frame_id,
+        annotation_revision="sha256:" + "0" * 64,
+        annotations=built.annotations,
+    )
+    forged.document_hash()  # caching the hash must not mark the document verified
+    with pytest.raises(AnnotationContractError) as error:
+        parse_clip_annotations(forged)
+    assert error.value.code == "annotation_revision_mismatch"
+
+    replaced = dataclasses.replace(built, annotation_revision="sha256:" + "1" * 64)
+    with pytest.raises(AnnotationContractError):
+        parse_clip_annotations(replaced)