Defined in awkward.operations.convert on line 3849.

ak.from_parquet(source, columns=None, row_groups=None, use_threads=True, include_partition_columns=True, lazy=False, lazy_cache='new', lazy_cache_key=None, highlevel=True, behavior=None)
  • source (str, Path, file-like object, pyarrow.NativeFile) – Where to get the Parquet file. If source is the name of a local directory (str or Path), then it is interpreted as a partitioned Parquet dataset.

  • columns (None or list of str) – If None, read all columns; otherwise, read a specified set of columns.

  • row_groups (None, int, or list of int) – If None, read all row groups; otherwise, read a single or list of row groups.

  • use_threads (bool) – Passed to the pyarrow.parquet.ParquetFile.read functions; if True, do multithreaded reading.

  • include_partition_columns (bool) – If True and source is a partitioned Parquet dataset with subdirectory names defining partition names and values, include those special columns in the output.

  • lazy (bool) – If True, read columns in row groups on demand (as ak.layout.VirtualArray, possibly in ak.partition.PartitionedArray if the file has more than one row group); if False, read all requested data immediately.

  • lazy_cache (None, "new", or MutableMapping) – If lazy, pass this cache to the VirtualArrays. If “new”, a new dict (keep-forever cache) is created. If None, no cache is used.

  • lazy_cache_key (None or str) – If lazy, pass this cache_key to the VirtualArrays. If None, a process-unique string is constructed.

  • highlevel (bool) – If True, return an ak.Array; otherwise, return a low-level ak.layout.Content subclass.

  • behavior (None or dict) – Custom ak.behavior for the output array, if high-level.

  • options – All other options are passed to pyarrow.parquet.ParquetFile.

Reads a Parquet file into an Awkward Array (through pyarrow).

>>> ak.from_parquet("array1.parquet")
<Array [[1, 2, 3], [], ... [], [6, 7, 8, 9]] type='6 * var * ?int64'>

See also ak.from_arrow, which is used as an intermediate step. See also ak.to_parquet.