pymetagen 0.4.0

Metadata generator using polars as backend.

PyMetaGen

PyMetaGen is a powerful and fast data quality tool base on Polars designed for generating metadata and extracting useful information from various data file formats. It provides both a Python API and a command-line interface (CLI) to inspect, filter, and extract data from files such as CSV, JSON, Parquet, and Excel.

Key Features

• Metadata Generation: Automatically generates metadata for your datasets, including statistics such as min, max, standard deviation, and more.
• Data Extraction: Easily extract specific rows from your datasets using head, tail, or random sampling.
• Command Line Interface: Perform operations like metadata generation, data inspection, and filtering using an intuitive CLI.
• Multiple File Format Support: Import and export data in various formats, including CSV, Parquet, Excel, and JSON.
• SQL Query Support: Filter data using SQL queries directly on the command line.

Installation

To install the package, use the following command:

pip install pymetagen

Local Installation

To install the package locally, use the following command:

python -m pip install -U git+ssh://[e-mail]/itsbigspark/dotdda.git@dev/main

Usage

Python API

You can use the Python API to load a data file and generate metadata:

from pymetagen import MetaGen

# Create an instance of the MetaGen class reading a data file

metagen = MetaGen.from_path("tests/data/testdata.csv", loading_mode="eager")

# Display the first few rows of the data

metagen.data.head()

# Generate metadata and reset the index

metadata = metagen.compute_metadata().reset_index()

# Save the metadata to a file

metagen.write_metadata("tests/data/testdata_metadata.csv")

Command Line Interface

• Metadata Generation Generate metadata for a tabular data file:

$ metagen metadata -i tests/data/testdata.csv -o tests/data/testdata_metadata.csv

>>> Generating metadata for tests/data/testdata.csv...

• Data Inspection Inspect a data file (e.g., a partitioned Parquet file):

metagen inspect -i tests/data/input_ab_partition.parquet

• Data Filtering Filter a data set using an SQL query:

metagen filter -i tests/data/testdata.csv -q "SELECT * FROM data WHERE imdb_score > 9"

• Data Extraction Extract a specific number of rows from a data set:

$ metagen extracts -i tests/data/testdata.csv -o tests.csv -n 3

>>> Writing extract in: tests-head.csv
>>> Writing extract in: tests-tail.csv
>>> Writing extract in: tests-sample.csv

Available Output Formats

• CSV
• Parquet
• JSON
• Excel