Blaze provides a familiar interface to various computational resources.

Blaze extends the usability of NumPy and Pandas to distributed and out-of-core computing. Blaze provides an interface similar to that of the NumPy ND-Array or Pandas DataFrame but maps these familiar interfaces onto a variety of other computational engines like Postgres or Spark.


The core of Blaze consists of

  • A type system to express data types and layouts
  • A symbolic expression system
  • A set of interfaces to common data formats
  • A set of interfaces to computational engines

This architecture allows a single Blaze code to run against several computational backends. Blaze depends on and exposes the hard work of countless other projects.

Under Development

Please be aware that Blaze is under active development. The project has gone through many changes and this documentation has not been kept uniformly in sync.

Older Versions

Older versions of these documents can be found here.