mirror of
https://github.com/lancedb/lancedb.git
synced 2026-03-29 03:50:39 +00:00
0e486511fa
This hooks up a new writer implementation for the `add()` method. The main immediate benefit is it allows streaming requests to remote tables, and at the same time allowing retries for most inputs. In NodeJS, we always convert the data to `Vec<RecordBatch>`, so it's always retry-able. For Python, all are retry-able, except `Iterator` and `pa.RecordBatchReader`, which can only be consumed once. Some, like `pa.datasets.Dataset` are retry-able *and* streaming. A lot of the changes here are to make the new DataFusion write pipeline maintain the same behavior as the existing Python-based preprocessing, such as: * casting input data to target schema * rejecting NaN values if `on_bad_vectors="error"` * applying embedding functions. In future PRs, we'll enhance these by moving the embedding calls into DataFusion and making sure we parallelize them. See: https://github.com/lancedb/lancedb/issues/3048 --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
114 lines
3.3 KiB
Python
114 lines
3.3 KiB
Python
# SPDX-License-Identifier: Apache-2.0
|
|
# SPDX-FileCopyrightText: Copyright The LanceDB Authors
|
|
|
|
from functools import singledispatch
|
|
from typing import List, Optional, Tuple, Union
|
|
|
|
from lancedb.pydantic import LanceModel, model_to_dict
|
|
import pyarrow as pa
|
|
|
|
from ._lancedb import RecordBatchStream
|
|
|
|
|
|
class AsyncRecordBatchReader:
|
|
"""
|
|
An async iterator over a stream of RecordBatches.
|
|
|
|
Also allows access to the schema of the stream
|
|
"""
|
|
|
|
def __init__(
|
|
self,
|
|
inner: Union[RecordBatchStream, pa.Table],
|
|
max_batch_length: Optional[int] = None,
|
|
):
|
|
"""
|
|
|
|
Attributes
|
|
----------
|
|
schema : pa.Schema
|
|
The schema of the batches produced by the stream.
|
|
Accessing the schema does not consume any data from the stream
|
|
"""
|
|
if isinstance(inner, pa.Table):
|
|
self._inner = self._async_iter_from_table(inner, max_batch_length)
|
|
self.schema: pa.Schema = inner.schema
|
|
elif isinstance(inner, RecordBatchStream):
|
|
self._inner = inner
|
|
self.schema: pa.Schema = inner.schema
|
|
else:
|
|
raise TypeError("inner must be a RecordBatchStream or a Table")
|
|
|
|
async def read_all(self) -> List[pa.RecordBatch]:
|
|
"""
|
|
Read all the record batches from the stream
|
|
|
|
This consumes the entire stream and returns a list of record batches
|
|
|
|
If there are a lot of results this may consume a lot of memory
|
|
"""
|
|
return [batch async for batch in self]
|
|
|
|
def __aiter__(self):
|
|
return self
|
|
|
|
async def __anext__(self) -> pa.RecordBatch:
|
|
return await self._inner.__anext__()
|
|
|
|
@staticmethod
|
|
async def _async_iter_from_table(
|
|
table: pa.Table, max_batch_length: Optional[int] = None
|
|
):
|
|
"""
|
|
Create an AsyncRecordBatchReader from a Table
|
|
|
|
This is useful when you have a Table that you want to iterate
|
|
over asynchronously
|
|
"""
|
|
batches = table.to_batches(max_chunksize=max_batch_length)
|
|
for batch in batches:
|
|
yield batch
|
|
|
|
|
|
def peek_reader(
|
|
reader: pa.RecordBatchReader,
|
|
) -> Tuple[pa.RecordBatch, pa.RecordBatchReader]:
|
|
if not isinstance(reader, pa.RecordBatchReader):
|
|
raise TypeError("reader must be a RecordBatchReader")
|
|
batch = reader.read_next_batch()
|
|
|
|
def all_batches():
|
|
yield batch
|
|
yield from reader
|
|
|
|
return batch, pa.RecordBatchReader.from_batches(batch.schema, all_batches())
|
|
|
|
|
|
@singledispatch
|
|
def to_arrow(data) -> pa.Table:
|
|
"""Convert a single data object to a pa.Table."""
|
|
raise NotImplementedError(f"to_arrow not implemented for type {type(data)}")
|
|
|
|
|
|
@to_arrow.register(pa.RecordBatch)
|
|
def _arrow_from_batch(data: pa.RecordBatch) -> pa.Table:
|
|
return pa.Table.from_batches([data])
|
|
|
|
|
|
@to_arrow.register(pa.Table)
|
|
def _arrow_from_table(data: pa.Table) -> pa.Table:
|
|
return data
|
|
|
|
|
|
@to_arrow.register(list)
|
|
def _arrow_from_list(data: list) -> pa.Table:
|
|
if not data:
|
|
raise ValueError("Cannot create table from empty list without a schema")
|
|
|
|
if isinstance(data[0], LanceModel):
|
|
schema = data[0].__class__.to_arrow_schema()
|
|
dicts = [model_to_dict(d) for d in data]
|
|
return pa.Table.from_pylist(dicts, schema=schema)
|
|
|
|
return pa.Table.from_pylist(data)
|