pandas.DataFrame.to_parquet#

DataFrame.to_parquet(path=None, *, engine='auto', compression='snappy', index=None, partition_cols=None, storage_options=None, **kwargs)[source]#

Write a DataFrame to the binary parquet format.

This function writes the dataframe as a parquet file. You can choose different parquet backends, and have the option of compression. See the user guide for more details.

Parameters:

pathstr, path object, file-like object, or None, default None: String, path object (implementing os.PathLike[str]), or file-like object implementing a binary write() function. If None, the result is returned as bytes. If a string or path, it will be used as Root Directory path when writing a partitioned dataset.
engine{‘auto’, ‘pyarrow’, ‘fastparquet’}, default ‘auto’: Parquet library to use. If ‘auto’, then the option io.parquet.engine is used. The default io.parquet.engine behavior is to try ‘pyarrow’, falling back to ‘fastparquet’ if ‘pyarrow’ is unavailable.
compressionstr or None, default ‘snappy’: Name of the compression to use. Use None for no compression. Supported options: ‘snappy’, ‘gzip’, ‘brotli’, ‘lz4’, ‘zstd’.
indexbool, default None: If True, include the dataframe’s index(es) in the file output. If False, they will not be written to the file. If None, similar to True the dataframe’s index(es) will be saved. However, instead of being saved as values, the RangeIndex will be stored as a range in the metadata so it doesn’t require much space and is faster. Other indexes will be included as columns in the file output.
partition_colslist, optional, default None: Column names by which to partition the dataset. Columns are partitioned in the order they are given. Must be None if path is not a string.
storage_optionsdict, optional: Extra options that make sense for a particular storage connection, e.g. host, port, username, password, etc. For HTTP(S) URLs the key-value pairs are forwarded to urllib.request.Request as header options. For other URLs (e.g. starting with “s3://”, and “gcs://”) the key-value pairs are forwarded to fsspec.open. Please see fsspec and urllib for more details, and for more examples on storage options refer here.
**kwargs: Additional arguments passed to the parquet library. See pandas io for more details.

Returns:

bytes if no path argument is provided else None: Returns the DataFrame converted to the binary parquet format as bytes if no path argument. Returns None and writes the DataFrame to the specified location in the Parquet format if the path argument is provided.

See also

read_parquet: Read a parquet file.
DataFrame.to_orc: Write an orc file.
DataFrame.to_csv: Write a csv file.
DataFrame.to_sql: Write to a sql table.
DataFrame.to_hdf: Write to hdf.

Notes

This function requires either the fastparquet or pyarrow library.
When saving a DataFrame with categorical columns to parquet, the file size may increase due to the inclusion of all possible categories, not just those present in the data. This behavior is expected and consistent with pandas’ handling of categorical data. To manage file size and ensure a more predictable roundtrip process, consider using Categorical.remove_unused_categories() on the DataFrame before saving.

Examples

>>> df = pd.DataFrame(data={"col1": [1, 2], "col2": [3, 4]})
>>> df.to_parquet("df.parquet.gzip", compression="gzip")  
>>> pd.read_parquet("df.parquet.gzip")  
   col1  col2
0     1     3
1     2     4

If you want to get a buffer to the parquet content you can use a io.BytesIO object, as long as you don’t use partition_cols, which creates multiple files.

>>> import io
>>> f = io.BytesIO()
>>> df.to_parquet(f)
>>> f.seek(0)
0
>>> content = f.read()