Function from_blocks

python/ray/data/read_api.py:142–162

Create a :class:`~ray.data.Dataset` from a list of blocks. This method is primarily used for testing. Unlike other methods like :func:`~ray.data.from_pandas` and :func:`~ray.data.from_arrow`, this method gaurentees that it won't modify the number of blocks. Args: blocks: Li

(blocks: List[Block])

@DeveloperAPI
def from_blocks(blocks: List[Block]):
    """Create a :class:`~ray.data.Dataset` from a list of blocks.

    This method is primarily used for testing. Unlike other methods like
    :func:`~ray.data.from_pandas` and :func:`~ray.data.from_arrow`, this method
    gaurentees that it won't modify the number of blocks.

    Args:
        blocks: List of blocks to create the dataset from.

    Returns:
        A :class:`~ray.data.Dataset` holding the blocks.
    """
    block_refs = [ray.put(block) for block in blocks]
    meta_with_schema = [BlockMetadataWithSchema.from_block(block) for block in blocks]

    from_blocks_op = FromBlocks(block_refs, meta_with_schema)
    stats = DatasetStats(metadata={"FromBlocks": meta_with_schema}, parent=None)
    context = DataContext.get_current().copy()
    logical_plan = LogicalPlan(from_blocks_op, context)
    return MaterializedDataset(logical_plan, context, stats)


@PublicAPI