Skip to main content

Streaming reactive and dataflow graphs in Python

Project description

Python Data Streams

Build Status Coverage PyPI PyPI Binder

Tributary is a library for constructing dataflow graphs in python. Unlike many other DAG libraries in python (airflow, luigi, prefect, dagster, dask, kedro, etc), tributary is not designed with data/etl pipelines or scheduling in mind. Instead, tributary is more similar to libraries like mdf, pyungo, streamz, or pyfunctional, in that it is designed to be used as the implementation for a data model. One such example is the greeks library, which leverages tributary to build data models for options pricing.

Installation

Install with pip:

pip install tributary

or with conda:

conda install -c conda-forge tributary

or from source:

python setup.py install

Note: If installing from source or with pip, you'll also need Graphviz itself if you want to visualize the graph using the .graphviz() method.

Stream Types

Tributary offers several kinds of streams:

Streaming

These are synchronous, reactive data streams, built using asynchronous python generators. They are designed to mimic complex event processors in terms of event ordering.

Functional

These are functional streams, built by currying python functions (callbacks).

Lazy

These are lazily-evaluated python streams, where outputs are propogated only as inputs change. They are implemented as directed acyclic graphs.

Examples

  • Streaming: In this example, we construct a variety of forward propogating reactive graphs.
  • Lazy: In this example, we construct a variety of lazily-evaluated directed acyclic computation graphs.
  • Automatic Differentiation: In this example, we use tributary to perform automatic differentiation on both lazy and streaming graphs.

Graph Visualization

You can visualize the graph with Graphviz. All streaming and lazy nodes support a graphviz method.

Streaming and lazy nodes also support ipydagred3 for live update monitoring.

Streaming

Here green indicates executing, yellow indicates stalled for backpressure, and red indicates that StreamEnd has been propogated (e.g. stream has ended).

Lazy

Here green indicates executing, and red indicates that the node is dirty. Note the the determination if a node is dirty is also done lazily (we can check with isDirty whcih will update the node's graph state.

Catalog

See the CATALOG for a full list of functions, transforms, sources, and sinks.

Support / Contributors

Thanks to the following organizations for providing code or financial support.

Nemoulous

License

This software is licensed under the Apache 2.0 license. See the LICENSE file for details.

Alternatives

Here is an incomplete list of libraries which implement similar/overlapping functionality

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

tributary-0.2.0.tar.gz (83.7 kB view hashes)

Uploaded source

Built Distribution

tributary-0.2.0-py2.py3-none-any.whl (113.1 kB view hashes)

Uploaded py2 py3

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page