maxframe.liteframe.read_parquet#

maxframe.liteframe.read_parquet(path, columns=None, filters=None, default_index_type=None, storage_options=None, chunk_bytes='64M', merge_small_files=True, groups_as_chunks=True, session=None, **kwargs)[source]#

Read a parquet file into a LiteFrame with Arrow dtypes.

Local (file:// or bare path) files are read with pandas and wrapped via from_local_df(). Remote paths are read in chunks by the MaxFrame backend; when a session is available the schema is inferred remotely, otherwise the tiler infers types from the Parquet file’s built-in schema metadata at execution time.

Parameters:

path (str or list of str) – Path or list of paths to the parquet file(s) to read. Local paths and remote URIs are both supported.
columns (list of str, optional) – Column names to select. If None, all columns are read.
filters (list of tuple or list of list of tuple, optional) – Predicate filters to apply via parquet row-group pushdown.
default_index_type (str, optional) – Index type for the resulting LiteFrame. If None, a default is chosen by the backend.
storage_options (dict, optional) – Extra options for the storage backend (e.g. credentials).
chunk_bytes (str, default '64M') – Target chunk size in bytes for splitting remote files.
merge_small_files (bool, default True) – Whether to merge small files into larger chunks.
groups_as_chunks (bool, default True) – Whether to treat each parquet row group as a separate chunk.
session (Session, optional) – Session to use for remote schema inference and execution. If None, the default session is used when available.
**kwargs – Additional keyword arguments forwarded to the parquet reader.

Returns:

A lazy LiteFrame with Arrow dtypes representing the parquet data. Call .execute() to materialize the results.

Return type:

LiteFrame