Reading and writing data from Parquet

The Parquet file format provides columnar serialization for pandas DataFrames. It reads and writes DataFrames efficiently in terms of storage and performance and shares data across distributed systems without information loss. The Parquet file format does not support duplicate and numeric columns.

There are two engines used to read and write Parquet files in pandas: pyarrow and the fastparquet engine. pandas's default Parquet engine is pyarrow; if pyarrow is unavailable, then it uses fastparquet. In our example, we are using pyarrow. Let's install pyarrow using pip:

pip install pyarrow

You can also install the pyarrow engine in the Jupyter Notebook by putting an ! before the pip keyword. Here is an example:

!pip install pyarrow

Let's write a file using the pyarrow engine:

# Write to a parquet file.
df.to_parquet('employee.parquet', engine='pyarrow')

In the preceding code example, we have written the using to_parquet...