Skip to main content

Doing some data things in a memory efficient manner

Project description

How To Data

  1. Split data.

  2. Create a generator that will take the data as an iterator, yielding key,value pairs.

  3. Sort each list of key/value pairs by the key.

  4. Use heap to merge lists of key/value pairs by the key.

  5. Group key/value pairs by the key.

  6. Reduce each key grouped values to one value yielding a single key/value pair.

In lieu of a key, you may use a key function as long as it produces the same key throughout the map-sort-merge-group phases.

Split data

Use split_file to split up your data files or use split_csv_file to split up csv files which may have multi-line fields to ensure they are not broken up.:

import os

import karld

big_file_names = [
    "bigfile1.csv",
    "bigfile2.csv",
    "bigfile3.csv"
]

data_path = os.path.join('path','to','data', 'root')


def main():
    for filename in big_file_names:
        # Name the directory to write the split files into based
        # on the name of the file.
        out_dir = os.path.join(data_path, 'split_data', filename.replace('.csv', ''))

        # Split the file, with a default max_lines=200000 per shard of the file.
        karld.io.split_csv_file(os.path.join(data_path, filename), out_dir)


if __name__ == "__main__":
    main()

When you’re generating data and want to shard it out to files based on quantity, use one of the split output functions such as split_file_output_csv, split_file_output or split_file_output_json:

import os
import pathlib

import karld


def main():
    """
    Python 2 version
    """

    items = (str(x) + os.linesep for x in range(2000))

    out_dir = pathlib.Path('shgen')
    karld.io.ensure_dir(str(out_dir))

    karld.io.split_file_output('big_data', items, str(out_dir))


if __name__ == "__main__":
    main()

CSV serializable data:

import pathlib

import karld


def main():
    """
    From a source of data, shard it to csv files.
    """
    if karld.is_py3():
        third = chr
    else:
        third = unichr

    # Your data source
    items = ((x, x + 1, third(x + 10)) for x in range(2000))

    out_dir = pathlib.Path('shard_out_csv')

    karld.io.ensure_dir(str(out_dir))

    karld.io.split_file_output_csv('big_data.csv', items, str(out_dir))


if __name__ == "__main__":
    main()

Rows of json serializable data:

import pathlib

import karld


def main():
    """
    From a source of data, shard it to csv files.
    """
    if karld.is_py3():
        third = chr
    else:
        third = unichr

    # Your data source
    items = ((x, x + 1, third(x + 10)) for x in range(2000))

    out_dir = pathlib.Path('shard_out_json')

    karld.io.ensure_dir(str(out_dir))

    karld.io.split_file_output_json('big_data.json', items, str(out_dir))


if __name__ == "__main__":
    main()

Documentation

Read the docs: http://karld.readthedocs.org/en/latest/

Expanded Getting Started at http://karld.readthedocs.org/en/latest/getting-started.html.

More examples are documented at http://karld.readthedocs.org/en/latest/source/example.html. View the source of the example files, for examples…

Contributing:

Submit any issues or questions here: https://github.com/johnwlockwood/karl_data/issues.

Make pull requests to development branch of

https://github.com/johnwlockwood/karl_data.

Documentation is written in reStructuredText and currently uses the

Sphinx style for field lists http://sphinx-doc.org/domains.html#info-field-lists

Check out closed pull requests to see the flow of development, as almost every change to master is done via a pull request on GitHub. Code Reviews are welcome, even on merged Pull Requests. Feel free to ask questions about the code.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

karld-0.2.8.tar.gz (33.8 kB view hashes)

Uploaded Source

Built Distributions

karld-0.2.8.macosx-10.9-intel.exe (94.0 kB view hashes)

Uploaded Source

karld-0.2.8-py3.4.egg (62.1 kB view hashes)

Uploaded Source

karld-0.2.8-py2.7.egg (60.5 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page