feat: add timestamp and duration support in leaf data hashing (ITL-438)

- Add is_timestamp and is_duration checks to _element_size_for_type (8 bytes each)
- Add test_timestamp_types_in_schema and test_duration_types_in_schema
- Add TestTimestampDurationHashing golden parity tests (15 cases)
  matching Rust starfix byte-for-byte

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: kurodo3[bot]

src/starfix/arrow_digester.py

@@ -642,6 +642,10 @@ def _element_size_for_type(dt: pa.DataType) -> int | None:
         return 4
     if pa.types.is_time64(dt):
         return 8
+    if pa.types.is_timestamp(dt):
+        return 8  # int64 physical storage; unit/tz are schema metadata
+    if pa.types.is_duration(dt):
+        return 8  # int64 physical storage; unit is schema metadata
     if pa.types.is_decimal(dt):
         return dt.bit_width // 8
     if pa.types.is_fixed_size_binary(dt):

tests/test_arrow_digester.py

@@ -67,6 +67,34 @@ def test_struct_fields_sorted_in_schema(self):
         # a_field should appear before z_field in the Struct array
         assert s.index('"a_field"') < s.index('"z_field"')
 
+    def test_timestamp_types_in_schema(self):
+        schema = pa.schema([
+            pa.field("ts_utc_s",  pa.timestamp("s",  tz="UTC"), nullable=False),
+            pa.field("ts_utc_ms", pa.timestamp("ms", tz="UTC"), nullable=False),
+            pa.field("ts_utc_us", pa.timestamp("us", tz="UTC"), nullable=False),
+            pa.field("ts_utc_ns", pa.timestamp("ns", tz="UTC"), nullable=False),
+            pa.field("ts_naive_us", pa.timestamp("us"), nullable=False),
+        ])
+        s = _serialized_schema(schema)
+        assert '{"Timestamp":["Second","UTC"]}' in s
+        assert '{"Timestamp":["Millisecond","UTC"]}' in s
+        assert '{"Timestamp":["Microsecond","UTC"]}' in s
+        assert '{"Timestamp":["Nanosecond","UTC"]}' in s
+        assert '{"Timestamp":["Microsecond",null]}' in s
+
+    def test_duration_types_in_schema(self):
+        schema = pa.schema([
+            pa.field("dur_s",  pa.duration("s"),  nullable=False),
+            pa.field("dur_ms", pa.duration("ms"), nullable=False),
+            pa.field("dur_us", pa.duration("us"), nullable=False),
+            pa.field("dur_ns", pa.duration("ns"), nullable=False),
+        ])
+        s = _serialized_schema(schema)
+        assert '{"Duration":"Second"}' in s
+        assert '{"Duration":"Millisecond"}' in s
+        assert '{"Duration":"Microsecond"}' in s
+        assert '{"Duration":"Nanosecond"}' in s
+
 
 # ── Schema hashing (golden values from Rust) ──────────────────────────
 

tests/test_golden_parity.py

@@ -607,3 +607,101 @@ def test_batch_split_independence(self):
         hash_split = digester.finalize()
 
         assert hash_combined == hash_split
+
+
+# ---------------------------------------------------------------------------
+# Timestamp and Duration hashing — golden values match Rust (ITL-438)
+# ---------------------------------------------------------------------------
+
+
+class TestTimestampDurationHashing:
+    """Golden-hash parity for timestamp and duration types.
+
+    All expected values were generated by the Rust starfix test suite
+    (timestamp_array_hashing / duration_array_hashing in tests/arrow_digester.rs).
+    Array contents: [0, None, 1_000] as raw i64 values.
+    """
+
+    # -- Timestamp (tz-aware UTC) ------------------------------------------
+
+    def test_timestamp_second_utc(self):
+        arr = pa.array([0, None, 1_000], type=pa.timestamp("s", tz="UTC"))
+        assert ArrowDigester.hash_array(arr).hex() == "0000017cef6166a14741f94906bd3146ffa197d1b5756d594eb46304b678b21e44c40b"
+
+    def test_timestamp_millisecond_utc(self):
+        arr = pa.array([0, None, 1_000], type=pa.timestamp("ms", tz="UTC"))
+        assert ArrowDigester.hash_array(arr).hex() == "000001fd253a835218d5605ce8bb348699f973b93e29fbe21fe6e532468f988ece4ce5"
+
+    def test_timestamp_microsecond_utc(self):
+        arr = pa.array([0, None, 1_000], type=pa.timestamp("us", tz="UTC"))
+        assert ArrowDigester.hash_array(arr).hex() == "0000013af741ca0fe5bd48d0bd39e1993c60e7c032f23c4cf4be6c11cc4b50ce56013e"
+
+    def test_timestamp_nanosecond_utc(self):
+        arr = pa.array([0, None, 1_000], type=pa.timestamp("ns", tz="UTC"))
+        assert ArrowDigester.hash_array(arr).hex() == "000001f5b86904b4b7f2c6a18c7c4febda142bf622e4d01752f42b8f832c0a68a4d9c0"
+
+    # -- Timestamp (tz-naive) ----------------------------------------------
+
+    def test_timestamp_second_naive(self):
+        arr = pa.array([0, None, 1_000], type=pa.timestamp("s"))
+        assert ArrowDigester.hash_array(arr).hex() == "0000016ee5249d352c03cc820af39e98b7043993d467207cb6ca34c8d13cfacf141e30"
+
+    def test_timestamp_millisecond_naive(self):
+        arr = pa.array([0, None, 1_000], type=pa.timestamp("ms"))
+        assert ArrowDigester.hash_array(arr).hex() == "000001ef46619d2be8259335ed0943f42ea3ced19ff6d75471a64b2f03aea89afa4842"
+
+    def test_timestamp_microsecond_naive(self):
+        arr = pa.array([0, None, 1_000], type=pa.timestamp("us"))
+        assert ArrowDigester.hash_array(arr).hex() == "0000017e624eb2847f79f940d7123980b076aea1c1ecd3adaa34bdae24fa00f05afda3"
+
+    def test_timestamp_nanosecond_naive(self):
+        arr = pa.array([0, None, 1_000], type=pa.timestamp("ns"))
+        assert ArrowDigester.hash_array(arr).hex() == "000001f677ef0e42e9e40ea092471f82283212833ab6b4ff9bb45e86e88590f8d15796"
+
+    # -- Timestamp: unit and tz produce different hashes -------------------
+
+    def test_timestamp_units_differ(self):
+        values = [1_000, 2_000]
+        hashes = [
+            ArrowDigester.hash_array(pa.array(values, type=pa.timestamp("s"))).hex(),
+            ArrowDigester.hash_array(pa.array(values, type=pa.timestamp("ms"))).hex(),
+            ArrowDigester.hash_array(pa.array(values, type=pa.timestamp("us"))).hex(),
+            ArrowDigester.hash_array(pa.array(values, type=pa.timestamp("ns"))).hex(),
+        ]
+        assert len(set(hashes)) == 4, "all 4 timestamp units must produce distinct hashes"
+
+    def test_timestamp_tz_differs(self):
+        values = [1_000, 2_000]
+        naive = ArrowDigester.hash_array(pa.array(values, type=pa.timestamp("us"))).hex()
+        utc = ArrowDigester.hash_array(
+            pa.array(values, type=pa.timestamp("us", tz="UTC"))
+        ).hex()
+        assert naive != utc, "tz-naive and tz=UTC must produce different hashes"
+
+    # -- Duration ----------------------------------------------------------
+
+    def test_duration_second(self):
+        arr = pa.array([0, None, 1_000], type=pa.duration("s"))
+        assert ArrowDigester.hash_array(arr).hex() == "0000013540d23a4abf1dfbc939a9e5514be69e364cec466ef900fa35050dc3cf2994fa"
+
+    def test_duration_millisecond(self):
+        arr = pa.array([0, None, 1_000], type=pa.duration("ms"))
+        assert ArrowDigester.hash_array(arr).hex() == "000001564a87e1e07898af07a4a212e3db83f911ac287134f28c772bf5fb4da683402c"
+
+    def test_duration_microsecond(self):
+        arr = pa.array([0, None, 1_000], type=pa.duration("us"))
+        assert ArrowDigester.hash_array(arr).hex() == "0000017904faf3043cf870f0c139cba0282cdaf11c590aeb80f2a0d6b103893ba5da2a"
+
+    def test_duration_nanosecond(self):
+        arr = pa.array([0, None, 1_000], type=pa.duration("ns"))
+        assert ArrowDigester.hash_array(arr).hex() == "00000189a38c5a7adf4b19b3e0d467f042ab2389fed8ae5f1d2af8555dd6131478a49b"
+
+    def test_duration_units_differ(self):
+        values = [1_000, 2_000]
+        hashes = [
+            ArrowDigester.hash_array(pa.array(values, type=pa.duration("s"))).hex(),
+            ArrowDigester.hash_array(pa.array(values, type=pa.duration("ms"))).hex(),
+            ArrowDigester.hash_array(pa.array(values, type=pa.duration("us"))).hex(),
+            ArrowDigester.hash_array(pa.array(values, type=pa.duration("ns"))).hex(),
+        ]
+        assert len(set(hashes)) == 4, "all 4 duration units must produce distinct hashes"