Skip to main content

A kedro-plugin that adds caching to kedro pipelines

Project description

Kedro Cache

:warning: This plugin is still under active developement and not fully tested. Do not use this in any production systems. Please report any issues that you find.

📝 Description

kedro-cache is a kedro plugin that plugin that enables the caching of data sets. The advantage is that the data sets are loaded from data catalog and not recomputed if they have not changed. If the input data sets or code have changed, the outputs are recomputed and the data catalog is updated. This plugin works out of the box with any kedro project without having to change the code. The logic on how to determine if the cached data set in the catalog should be used is described in the flow chart below.

Caching Flowchart

Disclaimer: The caching strategy determines if a node function has changes by simply looking at the immediate function body. This does not take into account other things such as called function, global variable etc. that might also have changed.

🏆 Features

  • Caching of node outputs in catalog
  • No change to kedro project needed
  • Integration with kedro data catalog
  • Configuration via config.yml file

🏗 Installation

The plugin can be install with pip

pip install kedro-cache

🚀 Enable Caching

In the root directory of your kedro project, run

kedro cache init

This will create a new file cache.yml in the conf directory of your kedro project in which you can configure the kedro-cache module. Although this step is optional as the plugin comes with default configurations.

Next let's assume that you have the following kedro pipeline for which you want to add caching. There are two nodes. One that reads data from a input dataset, does some computations and writes it to a intermediate dataset and one that reads the data from the intermediate dataset and writes it to the output dataset.


def register_pipelines() -> Dict[str, Pipeline]:
    default_pipeline = pipeline(
                func=lambda x: x,
                func=lambda x: x,
    return {"__default__": default_pipeline}

In order to add logging we simply just have to register all used data sets in the data catalog. Because if the first node want to use the cached output instead of recalculating it, it need to load it from the data catalog. This is only possible if it was stored there.

# catalog.yml

  type: pandas.CSVDataSet
  filepath: input.csv

  type: pandas.CSVDataSet
  filepath: intermediate.csv

  type: pandas.CSVDataSet
  filepath: output.csv

And that was it. Just by adding all files to the catalog you enabled caching.

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

kedro_cache-0.1.1.tar.gz (11.7 kB view hashes)

Uploaded Source

Built Distribution

kedro_cache-0.1.1-py3-none-any.whl (12.9 kB view hashes)

Uploaded Python 3

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