Data Lake Utility

Package to manipulate data from/into Amazon S3 using partitions compatible with Apache Hadoop filesystem. At this moment, this package was conceived to handle JSON and Parquet formats. That being said, it expects a Pandas DataFrame.

Data will be written into Amazon S3 as a multi-line JSON or Apache Parquet, compressed as GZIP.

Features

Convert list of dictionaries...

[
    {"brand": "Ford","model": "Mustang","year": 1965},
    {"brand": "Pontiac","model": "GTO","year": 1964},
    {"brand": "Lamborghini","model": "Miura","year": 1966}
]

...to multi-line JSON compressed as GZIP

{"brand": "Ford","model": "Mustang","year": 1965}
{"brand": "Pontiac","model": "GTO","year": 1964}
{"brand": "Lamborghini","model": "Miura","year": 1966}

...or to Apache Parquet compressed as GZIP

How to use

This package does not slice data into partitions at the moment. You must handle slicing of data to write into partitions desired.

Check this example below:

from datalake_utils.utils import DataLake
import pandas

data = [
    {
        "brand": "Ford",
        "model": "Mustang",
        "year": 1965
    },
    {
        "brand": "Pontiac",
        "model": "GTO",
        "year": 1964
    },
    {
        "brand": "Lamborghini",
        "model": "Miura",
        "year": 1966
    }
]

datalake = DataLake(
    bucket_name="vehicles",
    schema="motor_vehicles",
    table="cars",
    partitions=[
        {
            "key": "decade",
            "value": "1960s"
        }
    ],
)

datalake.append_to_s3(data=pandas.DataFrame(data), file_format="json")

It will create an object into Amazon S3 with the following structure:

motor_vehicles/
`-- cars/
    `-- decade=1960s/
        `-- 2c0fea6c-444e-11ed-969f-acde48001122.json.gz

To read all files from partition, do:

retrieved_data = datalake.read_from_s3(file_format="json")

retrieved_data is a Pandas DataFrame object. It is not possible to read a specific file using this function.

To convert a Pandas DataFrame into a list of tuples containing a single JSON:

datalake.df_to_tuples(data=retrieved_data)

To delete all files from partition, do:

datalake.delete_from_s3()