Merge origin/main into gatekeeper/fix-2923-1

python/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-python"
-version = "0.37.1-beta.1"
+version = "0.38.0-beta.2"
 publish = false
 edition.workspace = true
 description = "Python bindings for LanceDB"
@@ -15,10 +15,10 @@ name = "_lancedb"
 crate-type = ["cdylib"]
 
 [dependencies]
-arrow = { version = "58.0.0", features = ["pyarrow"] }
-async-trait = "0.1"
-bytes = "1"
-lancedb = { path = "../rust/lancedb", default-features = false }
+arrow = { workspace = true, features = ["pyarrow"] }
+async-trait.workspace = true
+bytes.workspace = true
+lancedb.workspace = true
 datafusion-common.workspace = true
 lance-core.workspace = true
 lance-namespace.workspace = true
@@ -29,17 +29,17 @@ log.workspace = true
 # Maturin enables extension-module mode for Python builds. Keeping it out of
 # Cargo features lets Rust unit tests link against libpython.
 pyo3 = { version = "0.28", features = ["abi3-py310", "chrono"] }
-chrono = { version = "0.4", default-features = false, features = ["clock"] }
+chrono.workspace = true
 pyo3-async-runtimes = { version = "0.28", features = [
     "attributes",
     "tokio-runtime",
 ] }
-pin-project = "1.1.5"
+pin-project.workspace = true
 futures.workspace = true
-serde = "1"
-serde_json = "1"
+serde.workspace = true
+serde_json.workspace = true
 snafu.workspace = true
-tokio = { version = "1.40", features = ["sync", "rt-multi-thread"] }
+tokio.workspace = true
 libc = "0.2"
 
 [build-dependencies]

python/python/lancedb/__init__.py

@@ -12,6 +12,7 @@ __version__ = importlib.metadata.version("lancedb")
 
 from ._lancedb import connect as lancedb_connect
 from ._lancedb import FtsToken
+from ._lancedb import LsmWriteSpec
 from ._lancedb import tokenize as _tokenize
 from .common import URI, sanitize_uri
 from urllib.parse import urlparse
@@ -518,6 +519,7 @@ __all__ = [
     "Job",
     "LanceDBConnection",
     "LanceNamespaceDBConnection",
+    "LsmWriteSpec",
     "RemoteDBConnection",
     "Session",
     "Table",

python/python/lancedb/_lancedb.pyi

@@ -198,6 +198,9 @@ class Connection(object):
     async def drop_table(
         self, name: str, namespace_path: Optional[List[str]] = None
     ) -> None: ...
+    async def drop_table_async(
+        self, name: str, namespace_path: Optional[List[str]] = None
+    ) -> Job: ...
     async def drop_all_tables(
         self, namespace_path: Optional[List[str]] = None
     ) -> None: ...
@@ -335,6 +338,11 @@ class Table:
     ) -> list[FtsToken]: ...
     async def delete(self, filter: Union[str, PyExpr]) -> DeleteResult: ...
     async def add_columns(self, columns: list[tuple[str, str]]) -> AddColumnsResult: ...
+    async def add_computed_columns(
+        self, columns: list[tuple[str, str]]
+    ) -> AddColumnsResult: ...
+    async def refresh_column(self, column: str) -> RefreshColumnResult: ...
+    async def refresh_column_async(self, column: str) -> Job: ...
     async def add_columns_with_schema(self, schema: pa.Schema) -> AddColumnsResult: ...
     async def alter_columns(
         self, columns: list[dict[str, Any]]
@@ -680,6 +688,10 @@ class LsmWriteSpec:
 class AddColumnsResult:
     version: int
 
+class RefreshColumnResult:
+    rows_filled: int
+    version: int
+
 class AlterColumnsResult:
     version: int
 

python/python/lancedb/db.py

@@ -524,6 +524,12 @@ class DBConnection(EnforceOverrides):
             namespace_path = []
         raise NotImplementedError
 
+    def drop_table_async(
+        self, name: str, namespace_path: Optional[List[str]] = None
+    ) -> Job:
+        """Start dropping a table and return its cleanup job."""
+        raise NotImplementedError
+
     def rename_table(
         self,
         cur_name: str,
@@ -1186,6 +1192,20 @@ class LanceDBConnection(DBConnection):
             )
         )
 
+    @override
+    def drop_table_async(
+        self, name: str, namespace_path: Optional[List[str]] = None
+    ) -> Job:
+        """Start dropping a table and return its cleanup job.
+
+        The table may become unavailable before its data files are removed.
+        Call :meth:`Job.wait` to wait for cleanup to finish.
+        """
+        if namespace_path is None:
+            namespace_path = []
+        job = LOOP.run(self._conn.drop_table_async(name, namespace_path=namespace_path))
+        return Job(job if isinstance(job, AsyncJob) else AsyncJob(job))
+
     @override
     def drop_all_tables(self, namespace_path: Optional[List[str]] = None):
         if namespace_path is None:
@@ -1963,6 +1983,23 @@ class AsyncConnection(object):
             if f"Table '{name}' was not found" not in str(e):
                 raise e
 
+    async def drop_table_async(
+        self,
+        name: str,
+        *,
+        namespace_path: Optional[List[str]] = None,
+    ) -> AsyncJob:
+        """Start dropping a table and return its cleanup job.
+
+        The table may become unavailable before its data files are removed.
+        Await :meth:`AsyncJob.wait` to wait for cleanup to finish.
+        """
+        if namespace_path is None:
+            namespace_path = []
+        return AsyncJob(
+            await self._inner.drop_table_async(name, namespace_path=namespace_path)
+        )
+
     async def drop_all_tables(self, namespace_path: Optional[List[str]] = None):
         """Drop all tables from the database.
 

python/python/lancedb/namespace.py

@@ -49,6 +49,7 @@ from lancedb._lancedb import (
 )
 from lancedb.background_loop import LOOP
 from lancedb.db import AsyncConnection, DBConnection
+from lancedb.job import AsyncJob, Job
 from lance_namespace import (
     LanceNamespace,
     connect as namespace_connect,
@@ -624,6 +625,18 @@ class LanceNamespaceDBConnection(DBConnection):
             namespace_path = []
         LOOP.run(self._inner.drop_table(name, namespace_path=namespace_path))
 
+    @override
+    def drop_table_async(
+        self, name: str, namespace_path: Optional[List[str]] = None
+    ) -> Job:
+        """Start dropping a table and return its cleanup job."""
+        if namespace_path is None:
+            namespace_path = []
+        job = LOOP.run(
+            self._inner.drop_table_async(name, namespace_path=namespace_path)
+        )
+        return Job(job if isinstance(job, AsyncJob) else AsyncJob(job))
+
     @override
     def rename_table(
         self,
@@ -1134,6 +1147,14 @@ class AsyncLanceNamespaceDBConnection:
             namespace_path = []
         await self._inner.drop_table(name, namespace_path=namespace_path)
 
+    async def drop_table_async(
+        self, name: str, namespace_path: Optional[List[str]] = None
+    ) -> AsyncJob:
+        """Start dropping a table and return its cleanup job."""
+        if namespace_path is None:
+            namespace_path = []
+        return await self._inner.drop_table_async(name, namespace_path=namespace_path)
+
     async def rename_table(
         self,
         cur_name: str,

python/python/lancedb/query.py

@@ -2235,6 +2235,7 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
             reranker=self._reranker,
             limit=self._limit,
             with_row_ids=True,
+            offset=self._offset,
         )
         return self._finish_hybrid_results(results)
 
@@ -2256,6 +2257,7 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
         reranker,
         limit: int,
         with_row_ids: bool,
+        offset: Optional[int] = None,
     ) -> pa.Table:
         if norm == "rank":
             vector_results = LanceHybridQueryBuilder._rank(vector_results, "_distance")
@@ -2332,7 +2334,7 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
             score_i = results.column_names.index("_score")
             results = results.set_column(score_i, "_score", original_scores)
 
-        results = results.slice(length=limit)
+        results = results.slice(offset=offset or 0, length=limit)
 
         if not with_row_ids:
             results = results.drop(["_rowid"])
@@ -2679,8 +2681,12 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
 
         # Apply common configurations
         if self._limit:
-            self._vector_query.limit(self._limit)
-            self._fts_query.limit(self._limit)
+            # The final offset/limit window is sliced out of the combined,
+            # reranked results, so each sub-query must fetch enough rows to
+            # cover the skipped prefix as well as the window itself.
+            sub_query_limit = self._limit + (self._offset or 0)
+            self._vector_query.limit(sub_query_limit)
+            self._fts_query.limit(sub_query_limit)
         if self._columns:
             self._vector_query.select(self._columns)
             self._fts_query.select(self._columns)

python/python/lancedb/remote/db.py

@@ -23,7 +23,7 @@ import pyarrow as pa
 
 from ..common import DATA
 from ..db import DBConnection, LOOP
-from ..job import Job
+from ..job import AsyncJob, Job
 
 if TYPE_CHECKING:
     from .._lancedb import JobDescription, JobInfo
@@ -663,6 +663,16 @@ class RemoteDBConnection(DBConnection):
             namespace_path = []
         LOOP.run(self._conn.drop_table(name, namespace_path=namespace_path))
 
+    @override
+    def drop_table_async(
+        self, name: str, namespace_path: Optional[List[str]] = None
+    ) -> Job:
+        """Start dropping a table and return its cleanup job."""
+        if namespace_path is None:
+            namespace_path = []
+        job = LOOP.run(self._conn.drop_table_async(name, namespace_path=namespace_path))
+        return Job(job if isinstance(job, AsyncJob) else AsyncJob(job))
+
     @override
     def rename_table(
         self,

python/python/lancedb/remote/table.py

@@ -958,8 +958,19 @@ class RemoteTable(Table):
     def count_rows(self, filter: Optional[str] = None) -> int:
         return LOOP.run(self._table.count_rows(filter))
 
-    def add_columns(self, transforms: Dict[str, str]) -> AddColumnsResult:
-        return LOOP.run(self._table.add_columns(transforms))
+    def add_columns(
+        self,
+        transforms: Dict[str, str] | None = None,
+        *,
+        computed: Dict[str, str] | None = None,
+    ) -> AddColumnsResult:
+        return LOOP.run(self._table.add_columns(transforms, computed=computed))
+
+    def refresh_column(self, column: str):
+        return LOOP.run(self._table.refresh_column(column))
+
+    def refresh_column_async(self, column: str) -> Job:
+        return Job(LOOP.run(self._table.refresh_column_async(column)))
 
     def alter_columns(
         self, *alterations: Iterable[Dict[str, str]]
@@ -979,17 +990,39 @@ class RemoteTable(Table):
         return LOOP.run(self._table.set_unenforced_primary_key(columns))
 
     def set_lsm_write_spec(self, spec: "LsmWriteSpec") -> None:
-        """Not supported on LanceDB Cloud."""
+        """Install an LsmWriteSpec."""
         return LOOP.run(self._table.set_lsm_write_spec(spec))
 
     def unset_lsm_write_spec(self) -> None:
-        """Not supported on LanceDB Cloud."""
+        """Remove the LsmWriteSpec."""
         return LOOP.run(self._table.unset_lsm_write_spec())
 
     def get_lsm_write_spec(self) -> Optional["LsmWriteSpec"]:
         """Read the installed LsmWriteSpec, or ``None``."""
         return LOOP.run(self._table.get_lsm_write_spec())
 
+    def checkpoint_lsm(self) -> None:
+        """Synchronous version of
+        [`AsyncTable.checkpoint_lsm`][lancedb.AsyncTable.checkpoint_lsm]."""
+        return LOOP.run(self._table.checkpoint_lsm())
+
+    def flush_lsm(self) -> None:
+        """Synchronous version of
+        [`AsyncTable.flush_lsm`][lancedb.AsyncTable.flush_lsm]."""
+        return LOOP.run(self._table.flush_lsm())
+
+    def compact_lsm(self) -> None:
+        """Synchronous version of
+        [`AsyncTable.compact_lsm`][lancedb.AsyncTable.compact_lsm]."""
+        return LOOP.run(self._table.compact_lsm())
+
+    def get_lsm_stats(self, *, include_generation_rows: bool = False) -> Optional[dict]:
+        """Synchronous version of
+        [`AsyncTable.get_lsm_stats`][lancedb.AsyncTable.get_lsm_stats]."""
+        return LOOP.run(
+            self._table.get_lsm_stats(include_generation_rows=include_generation_rows)
+        )
+
     def close_lsm_writers(self) -> None:
         """No-op on LanceDB Cloud (no local shard writers)."""
         return LOOP.run(self._table.close_lsm_writers())

python/python/lancedb/streaming.py

@@ -11,6 +11,11 @@ Provides StreamingDataset, a PyTorch IterableDataset that guarantees:
 - **Resumability**: state_dict / load_state_dict capture per-split consumption
   counts so training can resume from an exact mid-epoch position even when the
   distributed topology changes between runs.
+
+Transform failures on bad rows (e.g. nulls or NaNs from incomplete data) can
+be tolerated with ``on_transform_error="skip"``; see the parameter
+documentation on StreamingDataset for how this interacts with the guarantees
+above.
 """
 
 import ctypes
@@ -22,7 +27,7 @@ import time
 from collections import deque
 from concurrent.futures import ThreadPoolExecutor
 from multiprocessing import RawArray
-from typing import Any, Callable, Iterator, Optional
+from typing import Any, Callable, Iterator, Optional, Union
 
 from torch.utils.data import IterableDataset, get_worker_info
 
@@ -127,6 +132,49 @@ class StreamingDataset(IterableDataset):
         Maximum number of transforms to run concurrently.  Must be greater
         than zero.  When ``None`` (the default), uses ``os.cpu_count()`` or 1
         when the CPU count is unavailable.
+    on_transform_error:
+        What to do when the transform raises an exception:
+
+        - ``"raise"`` (the default): the exception propagates and iteration
+          aborts.
+        - ``"skip"``: the failing rows are dropped and iteration continues.
+        - ``"warn"``: like ``"skip"``, but a warning is logged for each
+          failing batch.
+        - a callable ``handler(exc) -> bool``: called with the exception;
+          return ``True`` to skip the failing rows or ``False`` to re-raise.
+          Useful to skip only expected error types (compatible with
+          ``webdataset.handlers`` style handlers).
+
+        When a batch fails, the transform is re-invoked on each single-row
+        slice of the batch so that only the rows that actually fail are
+        dropped.  Transforms should therefore be deterministic and accept
+        batches of any size (including one row).  Skipped rows are counted in
+        ``rows_skipped``.
+
+        Skipping weakens the elastic-determinism guarantee at the end of the
+        epoch: splits that lose more rows than others run dry earlier, and
+        each rank's iterator ends at the last cycle where every split *it
+        owns* still has a row.  Because bad rows are not distributed evenly
+        across splits, this means one rank's iterator can yield noticeably
+        fewer or more steps than another rank's *in the same run* — there is
+        no cross-rank coordination that stops every rank at the same global
+        step.  This is generally safe for asynchronous or single-rank use,
+        but synchronous distributed training (e.g. ranks that call
+        ``all_reduce`` every step) can hang or deadlock if one rank's
+        iterator is exhausted while others are still stepping; callers doing
+        synchronous multi-rank training with ``on_transform_error != "raise"``
+        are responsible for their own cross-rank stopping mechanism (e.g.
+        broadcasting a stop signal on ``StopIteration``).  The final few
+        global steps can also differ across topologies (bounded by the skew
+        in bad-row counts across splits).  The sequence of samples yielded
+        from each split remains deterministic.  Mid-epoch
+        checkpoints remain exact provided the transform fails
+        deterministically; in multi-rank training each rank must save its
+        own ``state_dict`` and the states must be combined with
+        ``merge_state_dicts`` before resuming on a different topology.
+        Prefer the ``filter`` parameter when bad rows can be expressed as a
+        SQL predicate (e.g. ``"col IS NOT NULL"``) — filtering happens before
+        splits are built, so every guarantee is fully preserved.
     worker_info_override:
         If set, used in place of ``torch.utils.data.get_worker_info()`` to
         determine the DataLoader worker assignment.  Intended for unit tests
@@ -152,6 +200,7 @@ class StreamingDataset(IterableDataset):
         filter: Optional[str] = None,
         transform: Optional[Callable] = None,
         transform_parallelism: Optional[int] = None,
+        on_transform_error: Union[str, Callable[[Exception], bool]] = "raise",
         connection_factory: Optional[Callable[[str], Any]] = None,
         worker_info_override=None,
     ):
@@ -167,6 +216,13 @@ class StreamingDataset(IterableDataset):
             )
         if transform_parallelism is not None and transform_parallelism <= 0:
             raise ValueError("transform_parallelism must be greater than 0")
+        if on_transform_error not in ("raise", "skip", "warn") and not callable(
+            on_transform_error
+        ):
+            raise ValueError(
+                "on_transform_error must be 'raise', 'skip', 'warn', or a "
+                f"callable, got {on_transform_error!r}"
+            )
 
         self._table = table
         self._num_splits = num_splits
@@ -182,6 +238,7 @@ class StreamingDataset(IterableDataset):
         self._filter = filter
         self._transform = transform
         self._transform_parallelism = transform_parallelism
+        self._on_transform_error = on_transform_error
         self._connection_factory = connection_factory
         self._worker_info_override = worker_info_override
 
@@ -199,19 +256,28 @@ class StreamingDataset(IterableDataset):
         # in the main process.  RawArray is picklable via the forkserver
         # reduction protocol so it survives the dataset pickle round-trip.
         # Layout: [unscanned_rows, raw_rows, cooked_rows, consumed_rows,
-        #          bytes_loaded, fetch_time_us, transform_time_us]
-        self._worker_stats: RawArray = RawArray(ctypes.c_int64, 7)
+        #          bytes_loaded, fetch_time_us, transform_time_us,
+        #          rows_skipped]
+        self._worker_stats: RawArray = RawArray(ctypes.c_int64, 8)
 
         # Cumulative bytes of Arrow buffer data fetched across all iterations.
         self._bytes_loaded: int = 0
         # Cumulative seconds spent in LanceDB I/O and in transform functions.
         self._fetch_time: float = 0.0
         self._transform_time: float = 0.0
+        # Cumulative rows dropped by on_transform_error across all iterations.
+        self._rows_skipped: int = 0
 
         # Number of samples each split has already been consumed.  At global
         # step boundaries all splits have consumed this many samples, so a
         # single scalar captures the topology-independent checkpoint state.
         self._resume_offset: int = 0
+        # Permutation position each split has consumed through, keyed by
+        # global split index.  Equal to _resume_offset for every split unless
+        # on_transform_error skipped rows, in which case skipped positions
+        # push the watermark of the affected splits further ahead.  Splits
+        # this instance has never iterated have no entry.
+        self._resume_positions: dict[int, int] = {}
 
         # Build the permutation table once, deterministically.
         builder = permutation_builder(table)
@@ -275,6 +341,7 @@ class StreamingDataset(IterableDataset):
         # Set identity transform on each Permutation so __getitems__ returns
         # the raw RecordBatch.  Stage 2 applies the real transform.
         permutations: list[Permutation] = []
+        initial_positions: list[int] = []
         for split_idx in my_splits:
             perm = Permutation.from_tables(
                 self._table, self._perm_table, split=split_idx
@@ -282,14 +349,20 @@ class StreamingDataset(IterableDataset):
             if self._columns is not None:
                 perm = perm.select_columns(self._columns)
             perm = perm.with_transform(lambda batch: batch)
-            if self._resume_offset > 0:
-                perm = perm.with_skip(self._resume_offset)
+            start_pos = self._resume_positions.get(split_idx, self._resume_offset)
+            if start_pos > 0:
+                perm = perm.with_skip(start_pos)
+            initial_positions.append(start_pos)
             permutations.append(perm)
 
         n = len(permutations)
         split_sizes = [perm.num_rows for perm in permutations]
         initial_offset = self._resume_offset
         local_consumed = [0] * n
+        # Permutation position each split has consumed through (absolute,
+        # i.e. counted from the start of the unskipped split).  Runs ahead of
+        # initial + local_consumed when rows are skipped.
+        pos_consumed = list(initial_positions)
 
         batch_size = self._read_batch_size
         max_prefetch = self._prefetch_batches
@@ -302,12 +375,14 @@ class StreamingDataset(IterableDataset):
             self._transform if self._transform is not None else Transforms.arrow2python
         )
 
-        # Per-split pipeline state.
+        # Per-split pipeline state.  Batches are paired with the absolute
+        # permutation position of their first row so that skipped rows can be
+        # accounted for in pos_consumed.
         fetch_head = [0] * n
-        io_pending = [deque() for _ in range(n)]  # Future[RecordBatch]
-        raw_batches = [deque() for _ in range(n)]  # RecordBatch — fetched, awaiting tx
-        tx_pending = [deque() for _ in range(n)]  # Future[list[Any]]
-        cooked = [deque() for _ in range(n)]  # rows ready to yield
+        io_pending = [deque() for _ in range(n)]  # (abs_start, Future[RecordBatch])
+        raw_batches = [deque() for _ in range(n)]  # (abs_start, RecordBatch)
+        tx_pending = [deque() for _ in range(n)]  # Future[list[(abs_pos, row)]]
+        cooked = [deque() for _ in range(n)]  # (abs_pos, row) ready to yield
 
         # Limit simultaneous transforms to transform_workers across all splits.
         tx_semaphore = threading.Semaphore(transform_workers)
@@ -330,7 +405,8 @@ class StreamingDataset(IterableDataset):
             fetch_head[i] += fetch
             perm_i = permutations[i]
             indices = list(range(start, start + fetch))
-            io_pending[i].append(io_pool.submit(_io_call, perm_i, indices))
+            abs_start = initial_positions[i] + start
+            io_pending[i].append((abs_start, io_pool.submit(_io_call, perm_i, indices)))
 
         def _fill_io(i: int) -> None:
             while len(io_pending[i]) < max_prefetch and fetch_head[i] < split_sizes[i]:
@@ -338,15 +414,72 @@ class StreamingDataset(IterableDataset):
 
         def _drain_io(i: int) -> None:
             """Move completed I/O futures into raw_batches non-blockingly."""
-            while io_pending[i] and io_pending[i][0].done():
-                raw_batches[i].append(io_pending[i].popleft().result())
+            while io_pending[i] and io_pending[i][0][1].done():
+                abs_start, fut = io_pending[i].popleft()
+                raw_batches[i].append((abs_start, fut.result()))
 
         # ── Stage 2 helpers ───────────────────────────────────────────────────
 
-        def _tx_call_guarded(batch):
+        on_error = self._on_transform_error
+
+        def _should_skip(exc: Exception) -> bool:
+            if on_error == "raise":
+                return False
+            if callable(on_error):
+                return bool(on_error(exc))
+            return True  # "skip" or "warn"
+
+        def _check_row_count(rows: list, num_rows: int) -> None:
+            if len(rows) != num_rows:
+                raise ValueError(
+                    f"transform returned {len(rows)} rows for a batch of "
+                    f"{num_rows}; transforms must return exactly one output "
+                    "row per input row.  To drop bad rows, raise inside the "
+                    "transform and pass on_transform_error='skip'."
+                )
+
+        def _transform_isolated(abs_start, batch, batch_exc):
+            """Re-run the transform on single-row slices, dropping failures."""
+            out = []
+            skipped = 0
+            first_exc = None
+            for j in range(batch.num_rows):
+                try:
+                    rows = list(final_transform(batch.slice(j, 1)))
+                except Exception as exc:
+                    if not _should_skip(exc):
+                        raise
+                    skipped += 1
+                    if first_exc is None:
+                        first_exc = exc
+                    continue
+                _check_row_count(rows, 1)
+                out.append((abs_start + j, rows[0]))
+            self._rows_skipped += skipped
+            if skipped and on_error == "warn":
+                logger.warning(
+                    "Skipped %d of %d rows whose transform failed (first error: %r)",
+                    skipped,
+                    batch.num_rows,
+                    first_exc if first_exc is not None else batch_exc,
+                )
+            return out
+
+        def _transform_batch(abs_start, batch):
+            """Apply the transform, returning [(abs_pos, row), ...]."""
+            try:
+                rows = list(final_transform(batch))
+            except Exception as exc:
+                if not _should_skip(exc):
+                    raise
+                return _transform_isolated(abs_start, batch, exc)
+            _check_row_count(rows, batch.num_rows)
+            return [(abs_start + j, row) for j, row in enumerate(rows)]
+
+        def _tx_call_guarded(abs_start, batch):
             try:
                 t0 = time.perf_counter()
-                result = final_transform(batch)
+                result = _transform_batch(abs_start, batch)
                 self._transform_time += time.perf_counter() - t0
                 return result
             finally:
@@ -355,8 +488,8 @@ class StreamingDataset(IterableDataset):
         def _try_submit_tx(i: int) -> None:
             """Submit transforms for raw_batches[i] up to available capacity."""
             while raw_batches[i] and tx_semaphore.acquire(blocking=False):
-                batch = raw_batches[i].popleft()
-                tx_pending[i].append(tx_pool.submit(_tx_call_guarded, batch))
+                abs_start, batch = raw_batches[i].popleft()
+                tx_pending[i].append(tx_pool.submit(_tx_call_guarded, abs_start, batch))
 
         def _drain_tx(i: int) -> None:
             """Move completed transform futures into cooked non-blockingly."""
@@ -384,11 +517,14 @@ class StreamingDataset(IterableDataset):
                     # Acquire a transform slot (may block briefly if all
                     # transform_workers are busy with other splits).
                     tx_semaphore.acquire()
-                    batch = raw_batches[i].popleft()
-                    tx_pending[i].append(tx_pool.submit(_tx_call_guarded, batch))
+                    abs_start, batch = raw_batches[i].popleft()
+                    tx_pending[i].append(
+                        tx_pool.submit(_tx_call_guarded, abs_start, batch)
+                    )
                 elif io_pending[i]:
                     # Block on the oldest in-flight I/O fetch.
-                    raw_batches[i].append(io_pending[i].popleft().result())
+                    abs_start, fut = io_pending[i].popleft()
+                    raw_batches[i].append((abs_start, fut.result()))
                     _advance(i)
                 else:
                     break  # split exhausted
@@ -407,15 +543,28 @@ class StreamingDataset(IterableDataset):
                         _fill_io(i)
 
                     while True:
-                        # Stop when any split is exhausted (all exhaust
-                        # simultaneously: equal split sizes + round-robin).
-                        if any(local_consumed[i] >= split_sizes[i] for i in range(n)):
+                        # A cycle only runs if every split can still produce a
+                        # row.  Without skips all splits exhaust simultaneously
+                        # (equal split sizes + round-robin); when
+                        # on_transform_error drops rows a split can run dry
+                        # early, ending the epoch at the last complete cycle.
+                        # This check only sees splits owned by this rank/worker
+                        # (my_splits) — there is no cross-rank coordination, so
+                        # a different rank with fewer skipped rows keeps going;
+                        # see the on_transform_error docstring.
+                        exhausted = False
+                        for i in range(n):
+                            _ensure_cooked(i)
+                            if not cooked[i]:
+                                exhausted = True
+                                break
+                        if exhausted:
                             break
 
                         for i in range(n):
-                            _ensure_cooked(i)
-                            row = cooked[i].popleft()
+                            pos, row = cooked[i].popleft()
                             local_consumed[i] += 1
+                            pos_consumed[i] = pos + 1
                             _advance(i)
 
                             # After the last split in each cycle: update the
@@ -424,21 +573,39 @@ class StreamingDataset(IterableDataset):
                             # even when __iter__ runs in a worker process.
                             if i == n - 1:
                                 self._resume_offset = initial_offset + local_consumed[i]
+                                for j, split_idx in enumerate(my_splits):
+                                    self._resume_positions[split_idx] = pos_consumed[j]
                                 ws = self._worker_stats
                                 ws[0] = sum(
                                     split_sizes[j] - fetch_head[j] for j in range(n)
                                 )
                                 ws[1] = sum(
-                                    batch.num_rows for q in raw_batches for batch in q
+                                    batch.num_rows
+                                    for q in raw_batches
+                                    for _, batch in q
                                 )
                                 ws[2] = sum(len(q) for q in cooked)
                                 ws[3] = sum(local_consumed)
                                 ws[4] = self._bytes_loaded
                                 ws[5] = int(self._fetch_time * 1_000_000)
                                 ws[6] = int(self._transform_time * 1_000_000)
+                                ws[7] = self._rows_skipped
 
                             yield row
                 finally:
+                    # Final stats flush: the per-cycle write above never runs
+                    # when iteration ends mid-cycle (e.g. a split whose rows
+                    # were all skipped before completing a single cycle), so
+                    # counters like rows_skipped would otherwise be stale.
+                    ws = self._worker_stats
+                    ws[0] = sum(split_sizes[j] - fetch_head[j] for j in range(n))
+                    ws[1] = 0  # queue-depth properties document 0 when idle
+                    ws[2] = 0
+                    ws[3] = sum(local_consumed)
+                    ws[4] = self._bytes_loaded
+                    ws[5] = int(self._fetch_time * 1_000_000)
+                    ws[6] = int(self._transform_time * 1_000_000)
+                    ws[7] = self._rows_skipped
                     self._raw_batches_ref = None
                     self._cooked_ref = None
                     self._fetch_head_ref = None
@@ -492,7 +659,7 @@ class StreamingDataset(IterableDataset):
         batches.  Returns 0 when not iterating.
         """
         if self._raw_batches_ref is not None:
-            return sum(batch.num_rows for q in self._raw_batches_ref for batch in q)
+            return sum(batch.num_rows for q in self._raw_batches_ref for _, batch in q)
         return int(self._worker_stats[1])
 
     @property
@@ -522,6 +689,19 @@ class StreamingDataset(IterableDataset):
             )
         return int(self._worker_stats[0])
 
+    @property
+    def rows_skipped(self) -> int:
+        """Number of rows dropped because their transform raised an exception.
+
+        Only ever non-zero when ``on_transform_error`` is set to ``"skip"``,
+        ``"warn"``, or a callable that returned ``True``.  Accumulates across
+        multiple iterations of the same dataset instance and is never reset
+        automatically.
+        """
+        if self._raw_batches_ref is not None:
+            return self._rows_skipped
+        return int(self._worker_stats[7])
+
     @property
     def consumed_rows(self) -> int:
         """Number of rows already yielded to the caller across all splits.
@@ -587,12 +767,27 @@ class StreamingDataset(IterableDataset):
         every split has been consumed the same number of times (by the
         round-robin design), so the per-split count is a single uniform value
         that is identical across all ranks and DataLoader workers.
+
+        ``positions_consumed_per_split`` records how far into each split's
+        permutation iteration has advanced.  It only differs from
+        ``samples_consumed_per_split`` when ``on_transform_error`` skipped
+        rows, in which case entries are exact for the splits this instance
+        iterated and a lower bound (the sample count) for splits owned by
+        other ranks or workers.  Combine the state dicts from all ranks with
+        [merge_state_dicts][lancedb.streaming.StreamingDataset.merge_state_dicts]
+        to recover the exact value for every split before resuming on a
+        different topology.
         """
+        positions = [
+            self._resume_positions.get(split, self._resume_offset)
+            for split in range(self._num_splits)
+        ]
         return {
             "shuffle_seed": self._shuffle_seed,
             "num_splits": self._num_splits,
             "epoch": self._epoch,
             "samples_consumed_per_split": [self._resume_offset] * self._num_splits,
+            "positions_consumed_per_split": positions,
         }
 
     def load_state_dict(self, state: dict) -> None:
@@ -618,3 +813,96 @@ class StreamingDataset(IterableDataset):
             self._resume_offset = consumed[0] if consumed else 0
         else:
             self._resume_offset = int(consumed)
+        # Older checkpoints predate positions_consumed_per_split; without
+        # skipped rows positions equal sample counts, so falling back to
+        # _resume_offset (the .get default in __iter__) is exact.
+        positions = state.get("positions_consumed_per_split")
+        if positions is None:
+            self._resume_positions = {}
+        else:
+            self._resume_positions = {
+                split: int(pos) for split, pos in enumerate(positions)
+            }
+
+    @staticmethod
+    def merge_state_dicts(states: list[dict]) -> dict:
+        """Merge state dicts saved by different ranks into one exact state.
+
+        Only needed when ``on_transform_error`` skips rows in multi-rank
+        training: each rank then knows the exact permutation position only for
+        its own splits, and records a lower bound for the rest.  Because
+        exactly one rank owns each split, the elementwise maximum across all
+        ranks' ``positions_consumed_per_split`` recovers the exact position of
+        every split.  Without skipped rows every rank's state is already
+        identical and merging is a no-op.
+
+        Raises ``ValueError`` if the states are empty or were not produced by
+        the same run (mismatched seed, split count, epoch, or sample counts).
+
+        The merge is always all-to-all and topology-agnostic: collect the
+        ``state_dict()`` from every rank of the *previous* run into one list,
+        merge that whole list, and hand the identical merged result to every
+        rank of the *next* run — regardless of whether the rank count grew,
+        shrank, or stayed the same. There is no pairwise or subset merging
+        step, because each split's exact position is only known to whichever
+        rank owned that split, and the elementwise maximum needs every rank's
+        contribution to be correct.
+
+        For example, checkpointing 8 ranks and resuming on 4 (the same
+        pattern applies when growing, e.g. 4 ranks resuming on 8)::
+
+            states = [ds.state_dict() for ds in previous_run_datasets]  # 8
+            merged = StreamingDataset.merge_state_dicts(states)
+            for ds in resumed_datasets:  # now only 4 ranks
+                ds.load_state_dict(merged)  # same dict on every rank
+
+        The rank count on either side never affects the merge itself, since
+        ``merge_state_dicts`` only cares about the list of states it is
+        given.  Each split's position is recovered by elementwise maximum;
+        here rank 0 owned split 0 (and skipped two rows there) while rank 1
+        owned split 1 (and skipped one row):
+
+        >>> rank0 = {
+        ...     "shuffle_seed": 0, "num_splits": 2, "epoch": 0,
+        ...     "samples_consumed_per_split": [3, 3],
+        ...     "positions_consumed_per_split": [5, 3],
+        ... }
+        >>> rank1 = {
+        ...     "shuffle_seed": 0, "num_splits": 2, "epoch": 0,
+        ...     "samples_consumed_per_split": [3, 3],
+        ...     "positions_consumed_per_split": [3, 4],
+        ... }
+        >>> merged = StreamingDataset.merge_state_dicts([rank0, rank1])
+        >>> merged["positions_consumed_per_split"]
+        [5, 4]
+        """
+        if not states:
+            raise ValueError("merge_state_dicts requires at least one state dict")
+        first = states[0]
+        for state in states[1:]:
+            for key in ("shuffle_seed", "num_splits", "epoch"):
+                if state[key] != first[key]:
+                    raise ValueError(
+                        f"{key} mismatch across state dicts: "
+                        f"{state[key]} != {first[key]}"
+                    )
+            if (
+                state["samples_consumed_per_split"]
+                != first["samples_consumed_per_split"]
+            ):
+                raise ValueError(
+                    "samples_consumed_per_split mismatch across state dicts; "
+                    "state_dict() must be called at the same global step "
+                    "boundary on every rank"
+                )
+        merged = dict(first)
+        all_positions = [
+            state.get(
+                "positions_consumed_per_split", state["samples_consumed_per_split"]
+            )
+            for state in states
+        ]
+        merged["positions_consumed_per_split"] = [
+            max(per_split) for per_split in zip(*all_positions)
+        ]
+        return merged

python/python/lancedb/table.py

@@ -176,6 +176,7 @@ if TYPE_CHECKING:
         CompactionStats,
         Tag,
         AddColumnsResult,
+        RefreshColumnResult,
         AddResult,
         AlterColumnsResult,
         UpdateFieldMetadataResult,
@@ -432,6 +433,20 @@ def _cast_to_target_schema(
     return pa.RecordBatchReader.from_batches(reordered_schema, gen())
 
 
+def _field_extension_name(field: pa.Field) -> Optional[str]:
+    extension_name = getattr(field.type, "extension_name", None)
+    if extension_name is not None:
+        return extension_name
+
+    metadata = field.metadata or {}
+    extension_name = metadata.get(b"ARROW:extension:name") or metadata.get(
+        "ARROW:extension:name"
+    )
+    if isinstance(extension_name, bytes):
+        return extension_name.decode()
+    return extension_name
+
+
 def _align_field_types(
     fields: List[pa.Field],
     target_fields: List[pa.Field],
@@ -444,6 +459,16 @@ def _align_field_types(
         target_field = next((f for f in target_fields if f.name == field.name), None)
         if target_field is None:
             raise ValueError(f"Field '{field.name}' not found in target schema")
+        # Preserve arrow.json input until it reaches Lance. LanceDB exposes stored
+        # JSON columns as lance.json (JSONB-backed LargeBinary), but casting the
+        # input to that storage type here merely relabels the raw JSON bytes as
+        # JSONB. Lance must see arrow.json so it can perform the JSONB encoding.
+        if (
+            _field_extension_name(field) == "arrow.json"
+            and _field_extension_name(target_field) == "lance.json"
+        ):
+            new_fields.append(field)
+            continue
         if pa.types.is_struct(target_field.type):
             if pa.types.is_struct(field.type):
                 new_type = pa.struct(
@@ -1916,7 +1941,14 @@ class Table(ABC):
 
     @abstractmethod
     def add_columns(
-        self, transforms: Dict[str, str] | pa.Field | List[pa.Field] | pa.Schema
+        self,
+        transforms: Dict[str, str]
+        | pa.Field
+        | List[pa.Field]
+        | pa.Schema
+        | None = None,
+        *,
+        computed: Dict[str, str] | None = None,
     ):
         """
         Add new columns with defined values.
@@ -1930,11 +1962,95 @@ class Table(ABC):
             Alternatively, a pyarrow Field or Schema can be provided to add
             new columns with the specified data types. The new columns will
             be initialized with null values.
+        computed: Dict[str, str], optional
+            A map of column name to a SQL expression defining the column. The
+            column's type and inputs are derived from the expression, so no
+            data type is supplied.
+
+            Unlike ``transforms``, the expression is stored rather than
+            evaluated now: the column is committed with no values, and rows get
+            them from [`refresh_column`][lancedb.table.Table.refresh_column].
+            Declaring one therefore costs the same on a large table as on an
+            empty one.
+
+            A refresh does not revisit rows it has already filled, so mutating
+            an input leaves the value computed at fill time; recomputing means
+            dropping the column and declaring it again. While a declaration
+            reads a column, that column cannot be renamed, retyped or dropped.
+
+            On LanceDB Cloud and Enterprise the expression is planned by the
+            server, and the refresh runs as a server job -- see
+            [`refresh_column_async`][lancedb.table.Table.refresh_column_async].
+            Cannot be combined with ``transforms``.
 
         Returns
         -------
         AddColumnsResult
             version: the new version number of the table after adding columns.
+
+        Examples
+        --------
+        >>> import lancedb
+        >>> db = lancedb.connect("./.lancedb")
+        >>> table = db.create_table("computed_demo", [{"x": 1}, {"x": 2}])
+        >>> table.add_columns(computed={"doubled": "x * 2"})
+        AddColumnsResult(version=2)
+        >>> table.refresh_column("doubled")
+        RefreshColumnResult(rows_filled=2, version=3)
+        >>> table.to_arrow().sort_by("x").to_pandas()
+           x  doubled
+        0  1        2
+        1  2        4
+        """
+
+    @abstractmethod
+    def refresh_column(self, column: str) -> "RefreshColumnResult":
+        """
+        Fill the rows of a computed column that hold no value yet.
+
+        Declared with ``add_columns(computed=...)``, a column starts empty and
+        gets its values here. Rows appended since the last refresh are filled
+        by the next one; rows already filled are left as they are, so the call
+        is idempotent and does not observe a mutated input.
+
+        Local tables only: a remote refresh runs as a server job, through
+        [`refresh_column_async`][lancedb.table.Table.refresh_column_async].
+
+        Parameters
+        ----------
+        column: str
+            The name of the computed column to fill.
+
+        Returns
+        -------
+        RefreshColumnResult
+            rows_filled: the number of rows given a value.
+            version: the new version number of the table.
+        """
+
+    @abstractmethod
+    def refresh_column_async(self, column: str) -> Job:
+        """
+        Like :meth:`refresh_column`, but returns a handle to the refresh job
+        instead of blocking until it completes.
+
+        The job may already be complete when returned; callers must not assume
+        the column is filled until :meth:`Job.wait` returns. Invalid input --
+        an unknown column, or one that is not computed -- raises here rather
+        than failing the job. On local tables the job runs in-process; on
+        LanceDB Cloud and Enterprise it is the server's backfill job.
+
+        Examples
+        --------
+        >>> import lancedb
+        >>> db = lancedb.connect("./.lancedb")
+        >>> table = db.create_table("computed_job_demo", [{"x": 1}, {"x": 2}])
+        >>> table.add_columns(computed={"doubled": "x * 2"})
+        AddColumnsResult(version=2)
+        >>> job = table.refresh_column_async("doubled")
+        >>> job.wait()
+        >>> job.status()
+        'finished'
         """
 
     @abstractmethod
@@ -3939,9 +4055,28 @@ class LanceTable(Table):
         return LOOP.run(self._table.index_stats(index_name))
 
     def add_columns(
-        self, transforms: Dict[str, str] | pa.field | List[pa.field] | pa.Schema
+        self,
+        transforms: Dict[str, str]
+        | pa.field
+        | List[pa.field]
+        | pa.Schema
+        | None = None,
+        *,
+        computed: Dict[str, str] | None = None,
     ) -> AddColumnsResult:
-        return LOOP.run(self._table.add_columns(transforms))
+        return LOOP.run(self._table.add_columns(transforms, computed=computed))
+
+    def refresh_column(self, column: str) -> "RefreshColumnResult":
+        """Fill a computed column's unfilled rows. See
+        [`AsyncTable.refresh_column`][lancedb.AsyncTable.refresh_column]."""
+        return LOOP.run(self._table.refresh_column(column))
+
+    def refresh_column_async(self, column: str) -> Job:
+        """Fill a computed column's unfilled rows, returning a handle to the
+        refresh job. See
+        [`Table.refresh_column_async`][lancedb.table.Table.refresh_column_async].
+        """
+        return Job(LOOP.run(self._table.refresh_column_async(column)))
 
     def alter_columns(
         self, *alterations: Iterable[Dict[str, str]]
@@ -4690,7 +4825,7 @@ class AsyncTable:
 
         Examples
         --------
-        >>> from lancedb._lancedb import LsmWriteSpec
+        >>> from lancedb import LsmWriteSpec
         >>> # table.set_unenforced_primary_key("id")
         >>> # table.set_lsm_write_spec(LsmWriteSpec.bucket("id", 16))
         """
@@ -4735,7 +4870,7 @@ class AsyncTable:
         ``asyncio.wait_for`` for a wall-clock bound; abandoning it partway
         costs nothing.
         """
-        return await self._inner.checkpoint_lsm()
+        await self._inner.checkpoint_lsm()
 
     async def flush_lsm(self) -> None:
         """Seal every bucket's active memtable into L0.
@@ -4744,7 +4879,7 @@ class AsyncTable:
         `compact_lsm`. On a node that has not claimed this table, this claims
         it and replays its WAL log first.
         """
-        return await self._inner.flush_lsm()
+        await self._inner.flush_lsm()
 
     async def compact_lsm(self) -> None:
         """Trigger a background L0 to base compaction pass per bucket.
@@ -4753,7 +4888,7 @@ class AsyncTable:
         ``get_lsm_stats`` for progress, or use ``checkpoint_lsm`` to loop
         until the current L0 has reached base.
         """
-        return await self._inner.compact_lsm()
+        await self._inner.compact_lsm()
 
     async def get_lsm_stats(
         self, *, include_generation_rows: bool = False
@@ -5856,7 +5991,14 @@ class AsyncTable:
         return await self._inner.update(updates_sql, where)
 
     async def add_columns(
-        self, transforms: dict[str, str] | pa.field | List[pa.field] | pa.Schema
+        self,
+        transforms: dict[str, str]
+        | pa.field
+        | List[pa.field]
+        | pa.Schema
+        | None = None,
+        *,
+        computed: dict[str, str] | None = None,
     ) -> AddColumnsResult:
         """
         Add new columns with defined values.
@@ -5869,6 +6011,22 @@ class AsyncTable:
             each row in the table, and can reference existing columns.
             Alternatively, you can pass a pyarrow field or schema to add
             new columns with NULLs.
+        computed: Dict[str, str], optional
+            A map of column name to a SQL expression defining the column. The
+            column's type and inputs are derived from the expression.
+
+            Unlike ``transforms``, the expression is stored rather than
+            evaluated now: the column is committed with no values, and rows get
+            them from
+            [`refresh_column`][lancedb.table.AsyncTable.refresh_column].
+
+            A refresh does not revisit rows it has already filled, so mutating
+            an input leaves the value computed at fill time. While a
+            declaration reads a column, that column cannot be renamed, retyped
+            or dropped.
+
+            On LanceDB Cloud and Enterprise the expression is planned by
+            the server. Cannot be combined with ``transforms``.
 
         Returns
         -------
@@ -5882,11 +6040,71 @@ class AsyncTable:
             {isinstance(f, pa.Field) for f in transforms}
         ):
             transforms = pa.schema(transforms)
+        if computed:
+            if transforms:
+                raise ValueError(
+                    "add_columns cannot take both transforms and computed columns"
+                )
+            return await self._inner.add_computed_columns(list(computed.items()))
+        if transforms is None:
+            raise ValueError("add_columns requires transforms or computed columns")
         if isinstance(transforms, pa.Schema):
             return await self._inner.add_columns_with_schema(transforms)
         else:
             return await self._inner.add_columns(list(transforms.items()))
 
+    async def refresh_column(self, column: str) -> RefreshColumnResult:
+        """
+        Fill the rows of a computed column that hold no value yet.
+
+        Declared with ``add_columns(computed=...)``, a column starts empty and
+        gets its values here. Rows appended since the last refresh are filled
+        by the next one; rows already filled are left as they are, so the call
+        is idempotent and does not observe a mutated input.
+
+        Local tables only: a remote refresh runs as a server job, through
+        [`refresh_column_async`][lancedb.table.Table.refresh_column_async].
+
+        Parameters
+        ----------
+        column: str
+            The name of the computed column to fill.
+
+        Returns
+        -------
+        RefreshColumnResult
+            The number of rows filled and the new version of the table.
+        """
+        return await self._inner.refresh_column(column)
+
+    async def refresh_column_async(self, column: str) -> AsyncJob:
+        """
+        Like :meth:`refresh_column`, but returns a handle to the refresh job
+        instead of blocking until it completes.
+
+        The job may already be complete when returned; callers must not assume
+        the column is filled until :meth:`AsyncJob.wait` resolves. Invalid
+        input -- an unknown column, or one that is not computed -- raises here
+        rather than failing the job. On local tables the job runs
+        in-process; on LanceDB Cloud and Enterprise it is the server's
+        backfill job.
+
+        Examples
+        --------
+        >>> import asyncio
+        >>> import lancedb
+        >>> async def refresh_in_background():
+        ...     db = await lancedb.connect_async("./.lancedb")
+        ...     table = await db.create_table("computed_job_async_demo", [{"x": 1}])
+        ...     await table.add_columns(computed={"doubled": "x * 2"})
+        ...     job = await table.refresh_column_async("doubled")
+        ...     await job.wait()
+        ...     return await job.status()
+        >>> asyncio.run(refresh_in_background())
+        'finished'
+        """
+        return AsyncJob(await self._inner.refresh_column_async(column))
+
     async def alter_columns(
         self, *alterations: Iterable[dict[str, Any]]
     ) -> AlterColumnsResult:

python/python/tests/test_db.py

@@ -755,8 +755,7 @@ def test_delete_table(tmp_db: lancedb.DBConnection):
     assert tmp_db.table_names() == []
 
 
-@pytest.mark.asyncio
-async def test_delete_table_async(tmp_db: lancedb.DBConnection):
+def test_drop_table_async(tmp_db: lancedb.DBConnection):
     data = pd.DataFrame(
         {
             "vector": [[3.1, 4.1], [5.9, 26.5]],
@@ -772,7 +771,10 @@ async def test_delete_table_async(tmp_db: lancedb.DBConnection):
 
     assert tmp_db.table_names() == ["test"]
 
-    tmp_db.drop_table("test")
+    job = tmp_db.drop_table_async("test")
+    assert job.id is None
+    assert job.status() == "finished"
+    job.wait()
     assert tmp_db.table_names() == []
 
     tmp_db.create_table("test", data=data)
@@ -781,6 +783,17 @@ async def test_delete_table_async(tmp_db: lancedb.DBConnection):
     tmp_db.drop_table("does_not_exist", ignore_missing=True)
 
 
+@pytest.mark.asyncio
+async def test_drop_table_async_connection(tmp_db_async: lancedb.AsyncConnection):
+    await tmp_db_async.create_table("test", data=pa.table({"id": [1, 2]}))
+
+    job = await tmp_db_async.drop_table_async("test")
+    assert job.id is None
+    assert await job.status() == "finished"
+    await job.wait()
+    assert await tmp_db_async.table_names() == []
+
+
 def test_drop_database(tmp_db: lancedb.DBConnection):
     data = pd.DataFrame(
         {

python/python/tests/test_elastic_dataloader.py

@@ -1456,6 +1456,408 @@ def test_shuffle_clump_size_yields_all_rows(lance_table):
     )
 
 
+# ---------------------------------------------------------------------------
+# on_transform_error tests
+# ---------------------------------------------------------------------------
+
+
+class BadRowError(ValueError):
+    """Raised by the failing transforms below when a batch contains a bad id."""
+
+
+def _failing_transform(bad_ids: set):
+    """A transform that raises BadRowError whenever the batch has a bad id.
+
+    Raises on the full batch and on any single-row slice containing a bad id,
+    so per-row isolation drops exactly the bad rows.
+    """
+
+    def transform(batch: pa.RecordBatch) -> list:
+        ids = batch.column("id").to_pylist()
+        bad = sorted(set(ids) & bad_ids)
+        if bad:
+            raise BadRowError(f"bad ids in batch: {bad}")
+        return [{"id": i} for i in ids]
+
+    return transform
+
+
+def _sequential_split_members(table) -> list[list[int]]:
+    """Return each split's ids in yield order for shuffle=False.
+
+    With a single rank and no workers the round-robin yields one row per split
+    per cycle, so item k of a clean run belongs to split k % NUM_SPLITS.
+    """
+    ds = StreamingDataset(table, num_splits=NUM_SPLITS, shuffle=False)
+    members: list[list[int]] = [[] for _ in range(NUM_SPLITS)]
+    for k, row in enumerate(ds):
+        members[k % NUM_SPLITS].append(row["id"])
+    return members
+
+
+def test_on_transform_error_default_raises(lance_table):
+    """By default a transform exception propagates and aborts iteration."""
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle_seed=SHUFFLE_SEED,
+        transform=_failing_transform({7}),
+    )
+    with pytest.raises(BadRowError):
+        list(ds)
+
+
+def test_on_transform_error_invalid_value(lance_table):
+    with pytest.raises(ValueError, match="on_transform_error"):
+        StreamingDataset(lance_table, num_splits=NUM_SPLITS, on_transform_error="bogus")
+
+
+def test_on_transform_error_skip_drops_bad_rows(lance_table):
+    """With one bad row per split, 'skip' yields every good row exactly once
+    and counts the dropped rows in rows_skipped."""
+    members = _sequential_split_members(lance_table)
+    bad_ids = {members[i][4] for i in range(NUM_SPLITS)}
+
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error="skip",
+    )
+    assert ds.rows_skipped == 0
+
+    ids = [row["id"] for row in ds]
+
+    assert sorted(ids) == sorted(set(range(NUM_ROWS)) - bad_ids)
+    assert ds.rows_skipped == NUM_SPLITS
+
+
+def test_on_transform_error_skip_uneven_ends_at_last_complete_cycle(lance_table):
+    """When one split loses more rows than the others, the epoch ends at the
+    last cycle where every split still has a row — no crash, no bad rows, and
+    every step remains one sample per split."""
+    members = _sequential_split_members(lance_table)
+    bad_ids = set(members[0][:3])  # all 3 bad rows in split 0
+
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error="skip",
+    )
+    items = [row["id"] for row in ds]
+
+    rows_per_split = NUM_ROWS // NUM_SPLITS
+    expected_cycles = rows_per_split - len(bad_ids)
+    assert len(items) == expected_cycles * NUM_SPLITS
+    assert len(set(items)) == len(items), "duplicate samples yielded"
+    assert not set(items) & bad_ids, "a bad row was yielded"
+    # Split 0 contributed exactly its surviving rows, in order, one per cycle.
+    survivors = [i for i in members[0] if i not in bad_ids]
+    assert items[0::NUM_SPLITS] == survivors[:expected_cycles]
+
+
+def test_on_transform_error_warn_logs(lance_table, caplog):
+    """'warn' skips like 'skip' but logs a warning for the failing batch."""
+    members = _sequential_split_members(lance_table)
+    bad_ids = {members[i][3] for i in range(NUM_SPLITS)}
+
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error="warn",
+    )
+    with caplog.at_level(logging.WARNING, logger="lancedb.streaming"):
+        items = list(ds)
+
+    assert len(items) == NUM_ROWS - NUM_SPLITS
+    assert ds.rows_skipped == NUM_SPLITS
+    assert "Skipped" in caplog.text
+    assert "BadRowError" in caplog.text
+
+
+def test_on_transform_error_callable_selective(lance_table):
+    """A callable handler can skip expected errors and re-raise the rest."""
+    members = _sequential_split_members(lance_table)
+    bad_ids = {members[i][0] for i in range(NUM_SPLITS)}
+
+    handled: list[Exception] = []
+
+    def handler(exc: Exception) -> bool:
+        handled.append(exc)
+        return isinstance(exc, BadRowError)
+
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error=handler,
+    )
+    items = list(ds)
+    assert len(items) == NUM_ROWS - NUM_SPLITS
+    assert handled and all(isinstance(exc, BadRowError) for exc in handled)
+
+    def broken_transform(batch: pa.RecordBatch) -> list:
+        raise TypeError("boom")
+
+    ds2 = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=broken_transform,
+        on_transform_error=handler,
+    )
+    with pytest.raises(TypeError, match="boom"):
+        list(ds2)
+
+
+def test_transform_wrong_row_count_raises(lance_table):
+    """A transform that returns the wrong number of rows is an error even with
+    on_transform_error='skip' — silent shrinkage would corrupt accounting."""
+
+    def drops_rows(batch: pa.RecordBatch) -> list:
+        return batch.column("id").to_pylist()[:-1]
+
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle_seed=SHUFFLE_SEED,
+        transform=drops_rows,
+        on_transform_error="skip",
+    )
+    with pytest.raises(ValueError, match="one output row per input row"):
+        list(ds)
+
+
+def test_skip_deterministic_across_runs(lance_table):
+    """With a fixed seed, skipping produces the identical sample sequence on
+    every run — skips are data-dependent, not run-dependent."""
+    bad_ids = {5, 17, 46}
+
+    def run() -> tuple[list[int], int]:
+        ds = StreamingDataset(
+            lance_table,
+            num_splits=NUM_SPLITS,
+            shuffle_seed=SHUFFLE_SEED,
+            transform=_failing_transform(bad_ids),
+            on_transform_error="skip",
+        )
+        return [row["id"] for row in ds], ds.rows_skipped
+
+    ids_a, skipped_a = run()
+    ids_b, skipped_b = run()
+    assert ids_a == ids_b
+    assert skipped_a == skipped_b
+    assert not set(ids_a) & bad_ids
+
+
+def test_skip_elastic_det_across_world_sizes(lance_table):
+    """With equal bad-row counts per split, skipping preserves the full
+    elastic-determinism guarantee: identical global batches at every step for
+    every compatible world_size."""
+    members = _sequential_split_members(lance_table)
+    bad_ids = {members[i][6] for i in range(NUM_SPLITS)}
+
+    def collect(world_size: int) -> list[frozenset[int]]:
+        micro = GLOBAL_BATCH_SIZE // world_size
+        iters = [
+            iter(
+                StreamingDataset(
+                    lance_table,
+                    num_splits=NUM_SPLITS,
+                    shuffle=False,
+                    rank=rank,
+                    world_size=world_size,
+                    transform=_failing_transform(bad_ids),
+                    on_transform_error="skip",
+                )
+            )
+            for rank in range(world_size)
+        ]
+        _STOP = object()
+        batches: list[frozenset[int]] = []
+        while True:
+            step_samples: set[int] = set()
+            exhausted = 0
+            for it in iters:
+                for _ in range(micro):
+                    val = next(it, _STOP)
+                    if val is _STOP:
+                        exhausted += 1
+                        break
+                    step_samples.add(val["id"])
+            if exhausted == len(iters):
+                break
+            assert exhausted == 0, (
+                "Rank iterators exhausted at different steps despite equal "
+                "bad-row counts per split"
+            )
+            batches.append(frozenset(step_samples))
+        return batches
+
+    reference = collect(1)
+    assert len(reference) == NUM_ROWS // NUM_SPLITS - 1
+    for ws in (2, 3, 4):
+        assert collect(ws) == reference, f"world_size={ws} diverged"
+
+
+def test_resumability_with_skips_same_topology(lance_table):
+    """Checkpointing mid-epoch with skipped rows resumes exactly: no sample
+    repeated, no sample lost, skipped rows stay skipped."""
+    members = _sequential_split_members(lance_table)
+    # Uneven skips: positions diverge across splits (2 bad in split 0, 1 in
+    # split 5), which only a position-based checkpoint can resume exactly.
+    bad_ids = {members[0][2], members[0][3], members[5][7]}
+    kwargs = dict(
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error="skip",
+    )
+
+    reference = [row["id"] for row in StreamingDataset(lance_table, **kwargs)]
+    rows_per_split = NUM_ROWS // NUM_SPLITS
+    assert len(reference) == (rows_per_split - 2) * NUM_SPLITS
+
+    steps = 3
+    ds = StreamingDataset(lance_table, **kwargs)
+    it = iter(ds)
+    consumed = [next(it)["id"] for _ in range(steps * NUM_SPLITS)]
+    checkpoint = ds.state_dict()
+    it.close()
+
+    # Split 0 skipped positions 2 and 3 within its first 3 yields; split 5's
+    # bad row is beyond the checkpoint.  Everything else is at 3 = the sample
+    # count.
+    positions = checkpoint["positions_consumed_per_split"]
+    assert positions[0] == 5
+    assert positions[1:] == [3] * (NUM_SPLITS - 1)
+    assert checkpoint["samples_consumed_per_split"] == [3] * NUM_SPLITS
+
+    ds2 = StreamingDataset(lance_table, **kwargs)
+    ds2.load_state_dict(checkpoint)
+    resumed = [row["id"] for row in ds2]
+
+    assert consumed == reference[: steps * NUM_SPLITS]
+    assert resumed == reference[steps * NUM_SPLITS :]
+
+
+def test_resumability_with_skips_elastic_merge(lance_table):
+    """Elastic resume with skips: each rank's checkpoint knows exact positions
+    only for its own splits; merge_state_dicts recovers the global state, and
+    a run on a different world_size continues exactly."""
+    members = _sequential_split_members(lance_table)
+    # Bad rows early in split 0 (rank 0) and split 6 (rank 1 of a ws=2 run) so
+    # both ranks' position vectors diverge before the checkpoint.
+    bad_ids = {members[0][0], members[0][2], members[6][1]}
+    kwargs = dict(
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error="skip",
+    )
+
+    reference = [row["id"] for row in StreamingDataset(lance_table, **kwargs)]
+
+    steps = 3
+    world_size = 2
+    micro = GLOBAL_BATCH_SIZE // world_size
+    datasets = [
+        StreamingDataset(lance_table, rank=rank, world_size=world_size, **kwargs)
+        for rank in range(world_size)
+    ]
+    iters = [iter(ds) for ds in datasets]
+    seen: list[frozenset[int]] = []
+    for _ in range(steps):
+        step_samples = set()
+        for it in iters:
+            for _ in range(micro):
+                step_samples.add(next(it)["id"])
+        seen.append(frozenset(step_samples))
+    states = [ds.state_dict() for ds in datasets]
+    for it in iters:
+        it.close()
+
+    merged = StreamingDataset.merge_state_dicts(states)
+    expected_positions = [3] * NUM_SPLITS
+    expected_positions[0] = 5  # skipped positions 0 and 2
+    expected_positions[6] = 4  # skipped position 1
+    assert merged["positions_consumed_per_split"] == expected_positions
+
+    # The first 3 global batches match the world_size=1 reference.
+    ref_batches = [
+        frozenset(reference[s * NUM_SPLITS : (s + 1) * NUM_SPLITS])
+        for s in range(len(reference) // NUM_SPLITS)
+    ]
+    assert seen == ref_batches[:steps]
+
+    # Resume on world_size=1 from the merged state.
+    ds_resume = StreamingDataset(lance_table, **kwargs)
+    ds_resume.load_state_dict(merged)
+    resumed = [row["id"] for row in ds_resume]
+    assert resumed == reference[steps * NUM_SPLITS :]
+
+
+def test_rows_skipped_flushed_when_split_entirely_bad(lance_table):
+    """A split whose rows all fail never completes a cycle, so the epoch ends
+    immediately — but rows_skipped must still report the drops after the
+    iterator exits (the shared-memory counter is flushed on exhaustion)."""
+    members = _sequential_split_members(lance_table)
+    bad_ids = set(members[0])  # every row of split 0 is bad
+
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error="skip",
+    )
+    assert list(ds) == []
+    assert ds.rows_skipped == len(bad_ids)
+
+
+def test_merge_state_dicts_validates_consistency(lance_table):
+    ds = StreamingDataset(lance_table, num_splits=NUM_SPLITS, shuffle_seed=SHUFFLE_SEED)
+    state = ds.state_dict()
+    other = dict(state, shuffle_seed=SHUFFLE_SEED + 1)
+    with pytest.raises(ValueError, match="shuffle_seed mismatch"):
+        StreamingDataset.merge_state_dicts([state, other])
+    with pytest.raises(ValueError, match="at least one"):
+        StreamingDataset.merge_state_dicts([])
+
+
+def test_load_state_dict_without_positions_key(lance_table):
+    """Checkpoints from before positions_consumed_per_split existed still
+    resume exactly (positions equal sample counts when nothing is skipped)."""
+    reference = [
+        row["id"]
+        for row in StreamingDataset(
+            lance_table, num_splits=NUM_SPLITS, shuffle_seed=SHUFFLE_SEED
+        )
+    ]
+
+    steps = 4
+    ds = StreamingDataset(lance_table, num_splits=NUM_SPLITS, shuffle_seed=SHUFFLE_SEED)
+    it = iter(ds)
+    for _ in range(steps * NUM_SPLITS):
+        next(it)
+    checkpoint = ds.state_dict()
+    it.close()
+    del checkpoint["positions_consumed_per_split"]
+
+    ds2 = StreamingDataset(
+        lance_table, num_splits=NUM_SPLITS, shuffle_seed=SHUFFLE_SEED
+    )
+    ds2.load_state_dict(checkpoint)
+    resumed = [row["id"] for row in ds2]
+    assert resumed == reference[steps * NUM_SPLITS :]
+
+
 def test_num_splits_defaults_to_world_size(lance_table):
     """Omitting num_splits gives world_size splits (one per rank)."""
     ds = StreamingDataset(

python/python/tests/test_expr.py

@@ -632,3 +632,101 @@ class TestExprBytesIntegration:
             .to_arrow()
         )
         assert result.num_rows == 2
+
+
+# ── datetime / timezone integration for lit() (issue #3262) ──────────────────
+
+
+class TestExprDatetimeTimezoneIntegration:
+    """Integration coverage for lit(datetime) against table timestamp columns.
+
+    PyArrow stores naive timestamps as UTC wall-clock microseconds. Python's
+    datetime.timestamp() treats naive values as *local* time, which used to
+    shift lit(naive) by the host UTC offset and break equality filters on
+    non-UTC machines. These cases lock the expected semantics.
+    """
+
+    def test_both_naive_match(self, tmp_path):
+        """Table naive + lit naive with the same wall clock must match."""
+        db = lancedb.connect(str(tmp_path / "naive"))
+        ts = datetime(2024, 7, 1, 10, 0, 0)
+        table = db.create_table(
+            "t", [{"id": 1, "ts": ts}, {"id": 2, "ts": datetime(2024, 7, 2, 10, 0, 0)}]
+        )
+        result = table.search().where(col("ts") == lit(ts)).to_list()
+        assert len(result) == 1
+        assert result[0]["id"] == 1
+
+    def test_both_same_timezone_match(self, tmp_path):
+        """Table UTC + lit UTC for the same instant must match."""
+        db = lancedb.connect(str(tmp_path / "utc"))
+        ts = datetime(2024, 7, 1, 10, 0, 0, tzinfo=timezone.utc)
+        table = db.create_table(
+            "t",
+            pa.table(
+                {
+                    "id": [1, 2],
+                    "ts": pa.array(
+                        [ts, datetime(2024, 7, 2, 10, 0, 0, tzinfo=timezone.utc)],
+                        type=pa.timestamp("us", tz="UTC"),
+                    ),
+                }
+            ),
+        )
+        result = table.search().where(col("ts") == lit(ts)).to_list()
+        assert len(result) == 1
+        assert result[0]["id"] == 1
+
+    def test_different_timezones_same_instant(self, tmp_path):
+        """UTC table row equals lit of the same instant in a different zone."""
+        db = lancedb.connect(str(tmp_path / "diff_tz"))
+        ts_utc = datetime(2024, 7, 1, 10, 0, 0, tzinfo=timezone.utc)
+        # Same instant as 06:00 in UTC-4
+        ts_est = datetime(2024, 7, 1, 6, 0, 0, tzinfo=timezone(timedelta(hours=-4)))
+        table = db.create_table(
+            "t",
+            pa.table(
+                {
+                    "id": [1],
+                    "ts": pa.array([ts_utc], type=pa.timestamp("us", tz="UTC")),
+                }
+            ),
+        )
+        result = table.search().where(col("ts") == lit(ts_est)).to_list()
+        assert len(result) == 1
+        assert result[0]["id"] == 1
+
+    def test_table_tz_literal_naive(self, tmp_path):
+        """UTC table + naive lit uses wall-clock equality (10:00 == 10:00 UTC)."""
+        db = lancedb.connect(str(tmp_path / "tz_naive"))
+        ts_utc = datetime(2024, 7, 1, 10, 0, 0, tzinfo=timezone.utc)
+        ts_naive = datetime(2024, 7, 1, 10, 0, 0)
+        table = db.create_table(
+            "t",
+            pa.table(
+                {
+                    "id": [1],
+                    "ts": pa.array([ts_utc], type=pa.timestamp("us", tz="UTC")),
+                }
+            ),
+        )
+        result = table.search().where(col("ts") == lit(ts_naive)).to_list()
+        assert len(result) == 1
+        assert result[0]["id"] == 1
+
+    def test_table_naive_literal_aware(self, tmp_path):
+        """Naive table + UTC lit with the same wall clock must match."""
+        db = lancedb.connect(str(tmp_path / "naive_aware"))
+        ts_naive = datetime(2024, 7, 1, 10, 0, 0)
+        ts_utc = datetime(2024, 7, 1, 10, 0, 0, tzinfo=timezone.utc)
+        table = db.create_table("t", [{"id": 1, "ts": ts_naive}])
+        result = table.search().where(col("ts") == lit(ts_utc)).to_list()
+        assert len(result) == 1
+        assert result[0]["id"] == 1
+
+    def test_naive_lit_sql_is_wall_clock_not_local_shifted(self):
+        """Regression: naive lit must not apply the host local UTC offset."""
+        ts = datetime(2024, 7, 1, 10, 0, 0)
+        sql = lit(ts).to_sql()
+        # Must encode 10:00 wall clock, not 10:00+local_offset.
+        assert "2024-07-01 10:00:00" in sql

python/python/tests/test_hybrid_query.py

@@ -203,6 +203,31 @@ async def test_async_hybrid_query_default_limit(table: AsyncTable):
     assert texts.count("a") == 1
 
 
+def test_hybrid_query_offset(sync_table: Table):
+    # The offset window of a hybrid query must be a suffix of the same query
+    # run without an offset -- it must not be silently ignored.
+    full = (
+        sync_table.search(query_type="hybrid")
+        .vector([0.0, 0.4])
+        .text("dog")
+        .limit(4)
+        .with_row_id(True)
+        .to_arrow()
+    )
+    assert len(full) == 4
+
+    offset_result = (
+        sync_table.search(query_type="hybrid")
+        .vector([0.0, 0.4])
+        .text("dog")
+        .offset(2)
+        .limit(2)
+        .with_row_id(True)
+        .to_arrow()
+    )
+    assert offset_result["_rowid"].to_pylist() == full["_rowid"].to_pylist()[2:]
+
+
 def test_hybrid_query_minimum_nprobes_zero_raises(sync_table: Table):
     # minimum_nprobes(0) must raise the same validation error a plain vector
     # query raises, not silently no-op because 0 is falsy.

python/python/tests/test_remote_db.py

@@ -1133,6 +1133,131 @@ def test_stats():
         assert res == stats
 
 
+@contextlib.contextmanager
+def lsm_test_table(lsm_handler):
+    """A remote table whose LSM routes are served by ``lsm_handler``.
+
+    ``lsm_handler(request, route)`` is called for ``/v1/table/test/<route>/``
+    where route is one of flush_lsm, compact_lsm, get_lsm_stats, and is
+    responsible for writing the response.
+    """
+    routes = ("flush_lsm", "compact_lsm", "get_lsm_stats")
+
+    def handler(request):
+        match = re.fullmatch(r"/v1/table/test/(\w+)/", request.path)
+        route = match.group(1) if match else None
+        if route in routes:
+            lsm_handler(request, route)
+        elif route == "describe":
+            request.send_response(200)
+            request.send_header("Content-Type", "application/json")
+            request.end_headers()
+            request.wfile.write(b'{"version": 1, "schema": {"fields": []}}')
+        else:
+            request.send_response(404)
+            request.end_headers()
+
+    with mock_lancedb_connection(handler) as db:
+        yield db.open_table("test")
+
+
+def read_json_body(request):
+    content_len = int(request.headers.get("Content-Length"))
+    return json.loads(request.rfile.read(content_len))
+
+
+def send_json(request, payload, status=200):
+    request.send_response(status)
+    request.send_header("Content-Type", "application/json")
+    request.end_headers()
+    request.wfile.write(json.dumps(payload).encode())
+
+
+def test_get_lsm_stats_sync():
+    """The sync wrapper round-trips the server payload into a dict."""
+    bucket = {
+        "shard_id": "b0",
+        "status": "Active",
+        "writer_epoch": 3,
+        "manifest_version": 12,
+        "current_generation": 6,
+        "replay_after_wal_entry_position": 40,
+        "wal_entry_position_last_seen": 42,
+        "generations": [{"generation": 5, "bytes": 1024, "rows": 7}],
+        "compacting": False,
+        "memtables": [
+            {
+                "generation": 6,
+                "rows": 2,
+                "bytes": 64,
+                "batches": 1,
+                "indexes": ["vec_idx"],
+            }
+        ],
+    }
+    seen_bodies = []
+
+    def lsm_handler(request, route):
+        assert route == "get_lsm_stats"
+        seen_bodies.append(read_json_body(request))
+        send_json(request, {"lsm_stats": {"buckets": [bucket]}})
+
+    with lsm_test_table(lsm_handler) as table:
+        assert table.get_lsm_stats() == {"buckets": [bucket]}
+        # Off by default, and forwarded when asked for.
+        assert seen_bodies == [{"include_generation_rows": False}]
+        table.get_lsm_stats(include_generation_rows=True)
+        assert seen_bodies[-1] == {"include_generation_rows": True}
+
+
+def test_get_lsm_stats_sync_returns_none_when_lsm_disabled():
+    """A null envelope means the LSM write path is not enabled, not an error."""
+
+    def lsm_handler(request, route):
+        send_json(request, {"lsm_stats": None})
+
+    with lsm_test_table(lsm_handler) as table:
+        assert table.get_lsm_stats() is None
+
+
+def test_flush_and_compact_lsm_sync():
+    """Both are one-shot POSTs answered 202 with no body."""
+    called = []
+
+    def lsm_handler(request, route):
+        called.append(route)
+        request.send_response(202)
+        request.end_headers()
+
+    with lsm_test_table(lsm_handler) as table:
+        assert table.flush_lsm() is None
+        assert table.compact_lsm() is None
+        assert called == ["flush_lsm", "compact_lsm"]
+
+
+def test_checkpoint_lsm_sync():
+    """Seal, read the watermark, and return once L0 holds nothing.
+
+    The convergence loop itself is covered in Rust; this pins the sync
+    binding to the endpoints it drives.
+    """
+    called = []
+
+    def lsm_handler(request, route):
+        called.append(route)
+        if route == "get_lsm_stats":
+            # An empty L0 yields no target watermark, so the loop is done
+            # after the seal without ever polling compaction.
+            send_json(request, {"lsm_stats": {"buckets": []}})
+        else:
+            request.send_response(202)
+            request.end_headers()
+
+    with lsm_test_table(lsm_handler) as table:
+        assert table.checkpoint_lsm() is None
+        assert called == ["flush_lsm", "get_lsm_stats"]
+
+
 @contextlib.contextmanager
 def query_test_table(query_handler, *, server_version=Version("0.1.0")):
     def handler(request):

python/python/tests/test_table.py

@@ -2772,6 +2772,56 @@ async def test_merge_insert_async(mem_db_async: AsyncConnection):
     assert (await table.to_arrow()).sort_by("a") == expected
 
 
+@pytest.mark.skipif(not hasattr(pa, "json_"), reason="requires PyArrow JSON type")
+@pytest.mark.asyncio
+async def test_merge_insert_encodes_json(mem_db_async: AsyncConnection):
+    json_type = pa.json_()
+    schema = pa.schema([pa.field("id", pa.string()), pa.field("j", json_type)])
+
+    def json_table(rows):
+        json_values = pa.ExtensionArray.from_storage(
+            json_type,
+            pa.array([value for _, value in rows], type=json_type.storage_type),
+        )
+        return pa.Table.from_arrays(
+            [pa.array([row_id for row_id, _ in rows]), json_values], schema=schema
+        )
+
+    table = await mem_db_async.create_table("json_merge", schema=schema)
+    await table.add(json_table([("a", '{"k": 1}'), ("b", '{"k": 9}')]))
+
+    await (
+        table.merge_insert("id")
+        .when_matched_update_all()
+        .execute(json_table([("a", '{"k": 2}')]))
+    )
+
+    rows = sorted(await table.query().to_list(), key=lambda row: row["id"])
+    assert rows == [
+        {"id": "a", "j": '{"k":2}'},
+        {"id": "b", "j": '{"k":9}'},
+    ]
+    filtered = await table.query().where("json_extract(j, '$.k') = '2'").to_list()
+    assert filtered == [{"id": "a", "j": '{"k":2}'}]
+
+
+@pytest.mark.skipif(not hasattr(pa, "json_"), reason="requires PyArrow JSON type")
+@pytest.mark.asyncio
+async def test_add_sanitization_encodes_json(mem_db_async: AsyncConnection):
+    json_type = pa.json_()
+    schema = pa.schema([pa.field("id", pa.string()), pa.field("j", json_type)])
+    json_values = pa.ExtensionArray.from_storage(
+        json_type, pa.array(['{"k": 3}'], type=json_type.storage_type)
+    )
+    data = pa.Table.from_arrays([pa.array(["c"]), json_values], schema=schema)
+
+    table = await mem_db_async.create_table("json_add", schema=schema)
+    await table.add(data, on_bad_vectors="fill")
+
+    rows = await table.query().where("json_extract(j, '$.k') = '3'").to_list()
+    assert rows == [{"id": "c", "j": '{"k":3}'}]
+
+
 def test_create_with_embedding_function(mem_db: DBConnection):
     class MyTable(LanceModel):
         text: str
@@ -3854,3 +3904,65 @@ async def test_async_search_runs_embedding_on_dedicated_executor(
     assert all(name.startswith("lancedb-embedding") for name in captured_threads), (
         f"embedding ran off the dedicated executor: {captured_threads}"
     )
+
+
+def test_computed_column_declare_and_refresh(tmp_path):
+    db = lancedb.connect(tmp_path)
+    table = db.create_table("computed", [{"x": 1}, {"x": 2}])
+
+    table.add_columns(computed={"doubled": "x * 2"})
+    assert table.to_arrow()["doubled"].to_pylist() == [None, None]
+
+    result = table.refresh_column("doubled")
+    assert result.rows_filled == 2
+    assert sorted(table.to_arrow()["doubled"].to_pylist()) == [2, 4]
+
+    table.add([{"x": 5}])
+    assert table.refresh_column("doubled").rows_filled == 1
+    assert sorted(table.to_arrow()["doubled"].to_pylist()) == [2, 4, 10]
+
+
+def test_computed_column_rejects_transforms_and_computed_together(tmp_path):
+    db = lancedb.connect(tmp_path)
+    table = db.create_table("computed_mixed", [{"x": 1}])
+    with pytest.raises(ValueError):
+        table.add_columns({"a": "x + 1"}, computed={"b": "x * 2"})
+
+
+@pytest.mark.asyncio
+async def test_computed_column_async(tmp_path):
+    db = await lancedb.connect_async(tmp_path)
+    table = await db.create_table("computed_async", [{"x": 3}])
+
+    await table.add_columns(computed={"tripled": "x * 3"})
+    await table.refresh_column("tripled")
+
+    assert (await table.to_arrow())["tripled"].to_pylist() == [9]
+
+
+def test_refresh_column_async_returns_job(tmp_path):
+    db = lancedb.connect(tmp_path)
+    table = db.create_table("computed_job", [{"x": 1}, {"x": 2}])
+    table.add_columns(computed={"doubled": "x * 2"})
+
+    job = table.refresh_column_async("doubled")
+    assert job.id is None  # in-process jobs have no server id
+    job.wait()
+    assert job.status() == "finished"
+    assert sorted(table.to_arrow()["doubled"].to_pylist()) == [2, 4]
+
+    # Bad input raises at the call, not through the job.
+    with pytest.raises(Exception, match="not a computed column"):
+        table.refresh_column_async("x")
+
+
+@pytest.mark.asyncio
+async def test_refresh_column_async_job_async_table(tmp_path):
+    db = await lancedb.connect_async(tmp_path)
+    table = await db.create_table("computed_job_async", [{"x": 3}])
+    await table.add_columns(computed={"tripled": "x * 3"})
+
+    job = await table.refresh_column_async("tripled")
+    await job.wait()
+    assert await job.status() == "finished"
+    assert (await table.to_arrow())["tripled"].to_pylist() == [9]

python/src/connection.rs

@@ -346,6 +346,23 @@ impl Connection {
         })
     }
 
+    #[pyo3(signature = (name, namespace_path=None))]
+    pub fn drop_table_async(
+        self_: PyRef<'_, Self>,
+        name: String,
+        namespace_path: Option<Vec<String>>,
+    ) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.get_inner()?.clone();
+        let ns_path = namespace_path.unwrap_or_default();
+        future_into_py(self_.py(), async move {
+            inner
+                .drop_table_async(name, &ns_path)
+                .await
+                .infer_error()
+                .map(crate::job::Job::new)
+        })
+    }
+
     #[pyo3(signature = (namespace_path=None,))]
     pub fn drop_all_tables(
         self_: PyRef<'_, Self>,

python/src/expr.rs

@@ -191,8 +191,27 @@ pub fn expr_lit(value: Bound<'_, PyAny>) -> PyResult<PyExpr> {
     }
 
     // datetime.datetime is a subclass of datetime.date, so it must be checked first.
+    //
+    // Python's datetime.timestamp() treats *naive* datetimes as local wall time.
+    // PyArrow (and therefore Lance table storage) encodes naive timestamps as
+    // UTC wall-clock microseconds. Using .timestamp() for naive values therefore
+    // shifts the literal by the local UTC offset on non-UTC machines, so
+    // `col("ts") == lit(naive_dt)` fails against a table that holds the same
+    // naive value. Fix: treat naive datetimes as UTC wall clock (match Arrow);
+    // keep aware datetimes on the real .timestamp() path (correct epoch).
     if let Ok(dt) = value.cast::<PyDateTime>() {
-        let ts: f64 = dt.call_method0("timestamp")?.extract()?;
+        let ts: f64 = if dt.getattr("tzinfo")?.is_none() {
+            // Force UTC interpretation of the naive wall clock.
+            let utc = pyo3::types::PyModule::import(value.py(), "datetime")?
+                .getattr("timezone")?
+                .getattr("utc")?;
+            let kwargs = pyo3::types::PyDict::new(value.py());
+            kwargs.set_item("tzinfo", utc)?;
+            let aware = dt.call_method("replace", (), Some(&kwargs))?;
+            aware.call_method0("timestamp")?.extract()?
+        } else {
+            dt.call_method0("timestamp")?.extract()?
+        };
         let micros = (ts * 1_000_000.0).round() as i64;
         return Ok(PyExpr(df_lit(ScalarValue::TimestampMicrosecond(
             Some(micros),

python/src/index.rs

@@ -297,7 +297,7 @@ struct IvfHnswFlatParams {
     target_partition_size: Option<u32>,
 }
 
-#[pyclass(get_all)]
+#[pyclass(module = "lancedb._lancedb", get_all)]
 /// A description of an index currently configured on a column
 pub struct IndexConfig {
     /// The type of the index

python/src/lib.rs

@@ -16,7 +16,8 @@ use query::{FTSQuery, HybridQuery, Query, VectorQuery};
 use session::Session;
 use table::{
     AddColumnsResult, AddResult, AlterColumnsResult, DeleteResult, DropColumnsResult, FtsToken,
-    LsmWriteSpec, MergeResult, PyBlobFile, Table, UpdateFieldMetadataResult, UpdateResult,
+    LsmWriteSpec, MergeResult, PyBlobFile, RefreshColumnResult, Table, UpdateFieldMetadataResult,
+    UpdateResult,
 };
 
 pub mod arrow;
@@ -57,6 +58,7 @@ pub fn _lancedb(_py: Python, m: &Bound<'_, PyModule>) -> PyResult<()> {
     m.add_class::<VectorQuery>()?;
     m.add_class::<RecordBatchStream>()?;
     m.add_class::<AddColumnsResult>()?;
+    m.add_class::<RefreshColumnResult>()?;
     m.add_class::<AlterColumnsResult>()?;
     m.add_class::<UpdateFieldMetadataResult>()?;
     m.add_class::<AddResult>()?;

python/src/session.rs

@@ -11,7 +11,7 @@ use pyo3::{PyResult, pyclass, pymethods};
 /// Sessions allow you to configure cache sizes for index and metadata caches,
 /// which can significantly impact memory use and performance. They can
 /// also be re-used across multiple connections to share the same cache state.
-#[pyclass(from_py_object)]
+#[pyclass(module = "lancedb._lancedb", from_py_object)]
 #[derive(Clone)]
 pub struct Session {
     pub(crate) inner: Arc<LanceSession>,

python/src/table.rs

@@ -415,6 +415,32 @@ pub struct AddColumnsResult {
     pub version: u64,
 }
 
+#[pyclass(get_all, from_py_object)]
+#[derive(Clone, Debug)]
+pub struct RefreshColumnResult {
+    pub rows_filled: u64,
+    pub version: u64,
+}
+
+#[pymethods]
+impl RefreshColumnResult {
+    pub fn __repr__(&self) -> String {
+        format!(
+            "RefreshColumnResult(rows_filled={}, version={})",
+            self.rows_filled, self.version
+        )
+    }
+}
+
+impl From<lancedb::table::RefreshColumnResult> for RefreshColumnResult {
+    fn from(result: lancedb::table::RefreshColumnResult) -> Self {
+        Self {
+            rows_filled: result.rows_filled,
+            version: result.version,
+        }
+    }
+}
+
 #[pymethods]
 impl AddColumnsResult {
     pub fn __repr__(&self) -> String {
@@ -579,7 +605,7 @@ impl PyBlobFile {
     }
 }
 
-#[pyclass(get_all, from_py_object)]
+#[pyclass(module = "lancedb._lancedb", get_all, from_py_object)]
 #[derive(Clone, Debug)]
 pub struct FtsToken {
     pub text: String,
@@ -1510,6 +1536,40 @@ impl Table {
         })
     }
 
+    pub fn add_computed_columns(
+        self_: PyRef<'_, Self>,
+        columns: Vec<(String, String)>,
+    ) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.inner_ref()?.clone();
+        future_into_py(self_.py(), async move {
+            let mut builder = inner.add_columns();
+            for (name, expression) in columns {
+                builder = builder.computed(name, expression);
+            }
+            let result = builder.execute().await.infer_error()?;
+            Ok(AddColumnsResult::from(result))
+        })
+    }
+
+    pub fn refresh_column(self_: PyRef<'_, Self>, column: String) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.inner_ref()?.clone();
+        future_into_py(self_.py(), async move {
+            let result = inner.refresh_column(column).await.infer_error()?;
+            Ok(RefreshColumnResult::from(result))
+        })
+    }
+
+    pub fn refresh_column_async(
+        self_: PyRef<'_, Self>,
+        column: String,
+    ) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.inner_ref()?.clone();
+        future_into_py(self_.py(), async move {
+            let job = inner.refresh_column_async(column).await.infer_error()?;
+            Ok(crate::job::Job::new(job))
+        })
+    }
+
     pub fn add_columns_with_schema(
         self_: PyRef<'_, Self>,
         schema: PyArrowType<Schema>,