class ParquetOptions

ParquetOptions(batch_size: int = 65536, row_group_size: int = 1048576, data_page_size: int = 1048576, compression: str = 'snappy', compression_level: int | None = None, columns: tuple[str, ...] = (), partition_columns: tuple[str, ...] = (), row_groups: tuple[int, ...] = (), predicates: tuple[ParquetPredicate, ...] = (), use_threads: bool = False, pre_buffer: bool = False, dataset: bool = False, hive: bool = False, unify: bool = False, normalize: bool = False, atomic: bool = False, durable: bool = False, reconstruct: bool = False, append: bool = False, encryption_key: bytes | None = None)

Source: packages/irx/src/irx/parquet_options.py:64

Options copied into a native Parquet session at open time.

Empty columns/row_groups select all. Dataset evolution unions same-typed fields only. Atomic replacement and rewrite append are POSIX-local operations with advisory writer exclusion and optional fsync durability. Reconstruct preserves supported logical types over explicit physical storage for single files. Raw AES keys are excluded from repr; callers own key distribution and lifetime.

Attributes

batch_size — Type: int
row_group_size — Type: int
data_page_size — Type: int
compression — Type: str
compression_level — Type: int | None
columns — Type: tuple[str, Ellipsis]
partition_columns — Type: tuple[str, Ellipsis]
row_groups — Type: tuple[int, Ellipsis]
predicates — Type: tuple[ParquetPredicate, Ellipsis]
use_threads — Type: bool
pre_buffer — Type: bool
dataset — Type: bool
hive — Type: bool
unify — Type: bool
normalize — Type: bool
atomic — Type: bool
durable — Type: bool
reconstruct — Type: bool
append — Type: bool
encryption_key — Type: bytes | None

Methods

__init__ —
__post_init__ — Validate collection items and secrets before any native side effect.
native — Create and release a temporary checked native options owner.