bug: support lag for non-UTC datetime cursor fields

- Add `ensure_pendulum_datetime_non_utc` to parse datetime strings into non-UTC datetime objects.
- Add `_datetime_obj_to_str` to preserve the colon in the timezone when converting datetime objects back to strings.
- Skip writing back state if no valid rows are found for `last_value` in the transformer, which may otherwise cause incorrect behavior.

dlt/common/time.py

@@ -125,6 +125,22 @@ def ensure_pendulum_datetime(value: TAnyDateTime) -> pendulum.DateTime:
     raise TypeError(f"Cannot coerce {value} to a pendulum.DateTime object.")
 
 
+def ensure_pendulum_datetime_non_utc(value: TAnyDateTime) -> pendulum.DateTime:
+    if isinstance(value, datetime.datetime):
+        ret = pendulum.instance(value)
+        return ret
+    elif isinstance(value, datetime.date):
+        return pendulum.datetime(value.year, value.month, value.day)
+    elif isinstance(value, (int, float, str)):
+        result = _datetime_from_ts_or_iso(value)
+        if isinstance(result, datetime.time):
+            raise ValueError(f"Cannot coerce {value} to a pendulum.DateTime object.")
+        if isinstance(result, pendulum.DateTime):
+            return result
+        return pendulum.datetime(result.year, result.month, result.day)
+    raise TypeError(f"Cannot coerce {value} to a pendulum.DateTime object.")
+
+
 def ensure_pendulum_time(value: Union[str, datetime.time]) -> pendulum.Time:
     """Coerce a time value to a `pendulum.Time` object.
 
@@ -164,27 +180,27 @@ def detect_datetime_format(value: str) -> Optional[str]:
         ): "%Y-%m-%dT%H:%M:%S.%fZ",  # UTC with fractional seconds
         re.compile(
             r"^\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}\+\d{2}:\d{2}$"
-        ): "%Y-%m-%dT%H:%M:%S%z",  # Positive timezone offset
+        ): "%Y-%m-%dT%H:%M:%S%:z",  # Positive timezone offset
         re.compile(
             r"^\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}\+\d{4}$"
         ): "%Y-%m-%dT%H:%M:%S%z",  # Positive timezone without colon
         # Full datetime with fractional seconds and positive timezone offset
         re.compile(
             r"^\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}\.\d+\+\d{2}:\d{2}$"
-        ): "%Y-%m-%dT%H:%M:%S.%f%z",
+        ): "%Y-%m-%dT%H:%M:%S.%f%:z",
         re.compile(
             r"^\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}\.\d+\+\d{4}$"
         ): "%Y-%m-%dT%H:%M:%S.%f%z",  # Positive timezone without colon
         re.compile(
             r"^\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}-\d{2}:\d{2}$"
-        ): "%Y-%m-%dT%H:%M:%S%z",  # Negative timezone offset
+        ): "%Y-%m-%dT%H:%M:%S%:z",  # Negative timezone offset
         re.compile(
             r"^\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}-\d{4}$"
         ): "%Y-%m-%dT%H:%M:%S%z",  # Negative timezone without colon
         # Full datetime with fractional seconds and negative timezone offset
         re.compile(
             r"^\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}\.\d+-\d{2}:\d{2}$"
-        ): "%Y-%m-%dT%H:%M:%S.%f%z",
+        ): "%Y-%m-%dT%H:%M:%S.%f%:z",
         re.compile(
             r"^\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}\.\d+-\d{4}$"
         ): "%Y-%m-%dT%H:%M:%S.%f%z",  # Negative Timezone without colon

dlt/extract/incremental/__init__.py

@@ -1,6 +1,7 @@
 import os
 from datetime import datetime  # noqa: I251
 from typing import Generic, ClassVar, Any, Optional, Type, Dict, Union, Literal, Tuple
+
 from typing_extensions import get_args
 
 import inspect
@@ -560,8 +561,14 @@ def __call__(self, rows: TDataItems, meta: Any = None) -> Optional[TDataItems]:
         else:
             rows = self._transform_item(transformer, rows)
 
-        # write back state
-        self._cached_state["last_value"] = transformer.last_value
+        # Skip writing back state if no valid rows for last_value in transformer may be wrong
+        if rows is not None:
+            is_empty_pandas = (
+                pandas is not None and isinstance(rows, pandas.DataFrame) and rows.empty
+            )
+            if not is_empty_pandas or rows != []:
+                self._cached_state["last_value"] = transformer.last_value
+
         if not transformer.deduplication_disabled:
             # compute hashes for new last rows
             unique_hashes = set(

dlt/extract/incremental/lag.py

@@ -1,12 +1,27 @@
+import sys
 from datetime import datetime, timedelta, date  # noqa: I251
 from typing import Union
 
 from dlt.common import logger
-from dlt.common.time import ensure_pendulum_datetime, detect_datetime_format
+from dlt.common.time import detect_datetime_format, ensure_pendulum_datetime_non_utc
 
 from . import TCursorValue, LastValueFunc
 
 
+def _datatime_obj_to_str(datatime: Union[datetime, date], datetime_format: str) -> str:
+    if sys.version_info < (3, 12, 0) and "%:z" in datetime_format:
+        modified_format = datetime_format.replace("%:z", "%z")
+        datetime_str = datatime.strftime(modified_format)
+
+        timezone_part = datetime_str[-5:] if len(datetime_str) >= 5 else ""
+        if timezone_part.startswith(("-", "+")):
+            return f"{datetime_str[:-5]}{timezone_part[:3]}:{timezone_part[3:]}"
+
+        raise ValueError(f"Invalid timezone format in datetime string: {datetime_str}")
+
+    return datatime.strftime(datetime_format)
+
+
 def _apply_lag_to_value(
     lag: float, value: TCursorValue, last_value_func: LastValueFunc[TCursorValue]
 ) -> TCursorValue:
@@ -17,12 +32,12 @@ def _apply_lag_to_value(
     is_str = isinstance(value, str)
     value_format = detect_datetime_format(value) if is_str else None
     is_str_date = value_format in ("%Y%m%d", "%Y-%m-%d") if value_format else None
-    parsed_value = ensure_pendulum_datetime(value) if is_str else value
+    parsed_value = ensure_pendulum_datetime_non_utc(value) if is_str else value
 
     if isinstance(parsed_value, (datetime, date)):
         parsed_value = _apply_lag_to_datetime(lag, parsed_value, last_value_func, is_str_date)  # type: ignore[assignment]
         # go back to string or pass exact type
-        value = parsed_value.strftime(value_format) if value_format else parsed_value  # type: ignore[assignment]
+        value = _datatime_obj_to_str(parsed_value, value_format) if value_format else parsed_value  # type: ignore[assignment]
 
     elif isinstance(parsed_value, (int, float)):
         value = _apply_lag_to_number(lag, parsed_value, last_value_func)  # type: ignore[assignment]

tests/common/test_time.py

@@ -12,8 +12,10 @@
     datetime_to_timestamp,
     datetime_to_timestamp_ms,
     detect_datetime_format,
+    ensure_pendulum_datetime_non_utc,
 )
 from dlt.common.typing import TAnyDateTime
+from dlt.extract.incremental.lag import _datatime_obj_to_str
 
 
 def test_timestamp_within() -> None:
@@ -132,21 +134,21 @@ def test_datetime_to_timestamp_helpers(
     [
         ("2024-10-20T15:30:00Z", "%Y-%m-%dT%H:%M:%SZ"),  # UTC 'Z'
         ("2024-10-20T15:30:00.123456Z", "%Y-%m-%dT%H:%M:%S.%fZ"),  # UTC 'Z' with fractional seconds
-        ("2024-10-20T15:30:00+02:00", "%Y-%m-%dT%H:%M:%S%z"),  # Positive timezone offset
+        ("2024-10-20T15:30:00+02:00", "%Y-%m-%dT%H:%M:%S%:z"),  # Positive timezone offset
         ("2024-10-20T15:30:00+0200", "%Y-%m-%dT%H:%M:%S%z"),  # Positive timezone offset (no colon)
         (
             "2024-10-20T15:30:00.123456+02:00",
-            "%Y-%m-%dT%H:%M:%S.%f%z",
+            "%Y-%m-%dT%H:%M:%S.%f%:z",
         ),  # Positive timezone offset with fractional seconds
         (
             "2024-10-20T15:30:00.123456+0200",
             "%Y-%m-%dT%H:%M:%S.%f%z",
         ),  # Positive timezone offset with fractional seconds (no colon)
-        ("2024-10-20T15:30:00-02:00", "%Y-%m-%dT%H:%M:%S%z"),  # Negative timezone offset
+        ("2024-10-20T15:30:00-02:00", "%Y-%m-%dT%H:%M:%S%:z"),  # Negative timezone offset
         ("2024-10-20T15:30:00-0200", "%Y-%m-%dT%H:%M:%S%z"),  # Negative timezone offset (no colon)
         (
             "2024-10-20T15:30:00.123456-02:00",
-            "%Y-%m-%dT%H:%M:%S.%f%z",
+            "%Y-%m-%dT%H:%M:%S.%f%:z",
         ),  # Negative timezone offset with fractional seconds
         (
             "2024-10-20T15:30:00.123456-0200",
@@ -170,6 +172,28 @@ def test_detect_datetime_format(value, expected_format) -> None:
     assert ensure_pendulum_datetime(value) is not None
 
 
+@pytest.mark.parametrize(
+    "datetime_str, datetime_format, expected_value",
+    [
+        ("2024-10-20T15:30:00+02:00", "%Y-%m-%dT%H:%M:%S%:z", "2024-10-20T15:30:00+02:00"),
+        ("2024-10-20T15:30:00+0200", "%Y-%m-%dT%H:%M:%S%z", "2024-10-20T15:30:00+0200"),
+        (
+            "2024-10-20T15:30:00.123456-02:00",
+            "%Y-%m-%dT%H:%M:%S.%f%:z",
+            "2024-10-20T15:30:00.123456-02:00",
+        ),
+        (
+            "2024-10-20T15:30:00.123456-0200",
+            "%Y-%m-%dT%H:%M:%S.%f%z",
+            "2024-10-20T15:30:00.123456-0200",
+        ),
+    ],
+)
+def test_datatime_obj_to_str(datetime_str, datetime_format, expected_value) -> None:
+    datetime = ensure_pendulum_datetime_non_utc(datetime_str)
+    assert _datatime_obj_to_str(datetime, datetime_format) == expected_value
+
+
 @pytest.mark.parametrize(
     "value",
     [

tests/extract/test_incremental.py

@@ -606,6 +606,39 @@ def some_data(created_at=dlt.sources.incremental("created_at", initial_value)):
     assert s["last_value"] == initial_value + timedelta(minutes=4)
 
 
+@pytest.mark.parametrize("item_type", ALL_TEST_DATA_ITEM_FORMATS)
+def test_incremental_transform_return_empty_rows_with_lag(item_type: TestDataItemFormat) -> None:
+    @dlt.resource
+    def some_data(
+        created_at=dlt.sources.incremental(
+            "created_at", initial_value="2024-11-01T08:00:00+08:00", lag=3600
+        )
+    ):
+        yield from source_items
+
+    p = dlt.pipeline(pipeline_name=uniq_id())
+
+    first_run_data = [{"id": 1, "value": 10, "created_at": "2024-11-01T12:00:00+08:00"}]
+    source_items = data_to_item_format(item_type, first_run_data)
+
+    p.extract(some_data())
+    s = p.state["sources"][p.default_schema_name]["resources"]["some_data"]["incremental"][
+        "created_at"
+    ]
+
+    assert s["last_value"] == "2024-11-01T12:00:00+08:00"
+
+    second_run_data = [{"id": 1, "value": 10, "created_at": "2024-11-01T10:00:00+08:00"}]
+    source_items = data_to_item_format(item_type, second_run_data)
+
+    p.extract(some_data())
+    s = p.state["sources"][p.default_schema_name]["resources"]["some_data"]["incremental"][
+        "created_at"
+    ]
+
+    assert s["last_value"] == "2024-11-01T12:00:00+08:00"
+
+
 @pytest.mark.parametrize("item_type", ALL_TEST_DATA_ITEM_FORMATS)
 def test_descending_order_unique_hashes(item_type: TestDataItemFormat) -> None:
     """Resource returns items in descending order but using `max` last value function.