Input/output#

LiteFrame IO functions read and write tabular data without index-specific metadata. All read functions use Arrow dtypes (dtype_backend="pyarrow") by default. Write functions have no index or index_label parameters because LiteFrame has no index semantics.

Flat file#

read_csv(path[, usecols, dtype, names, sep, ...])

Read a CSV file into a LiteFrame with Arrow dtypes.

LiteFrame.to_csv(path[, sep, na_rep, ...])

Write LiteFrame to a comma-separated values (csv) file.

JSON#

read_json(path, *[, lines, dtype, usecols, ...])

Read a JSON file into a LiteFrame with Arrow dtypes.

LiteFrame.to_json(path, *[, lines, orient, ...])

Write a LiteFrame to JSON format (line-delimited by default).

MaxCompute#

read_odps_table(table_name[, partitions, ...])

Read an ODPS table into a LiteFrame with Arrow dtypes.

LiteFrame.to_odps_table(table[, partition, ...])

Write LiteFrame object into a MaxCompute (ODPS) table.

Parquet#

read_parquet(path[, columns, filters, ...])

Read a parquet file into a LiteFrame with Arrow dtypes.

LiteFrame.to_parquet(path[, engine, ...])

Write a LiteFrame to the binary parquet format, each chunk will be written to a Parquet file.

Lance#

read_lance(path[, version, asof, columns, ...])

Read a Lance dataset into a LiteFrame.

LiteFrame.to_lance(path, *[, mode, on, ...])

Write a LiteFrame to a Lance dataset.