Skip to content

Container for Samples

Cyril Fournier requested to merge feature-sample_container into develop

Description

This EP introduces a new container class Sample for handling samples. The class is a wrapper around a pandas.DataFrame.
Several accessors give access to container content as numpy arrays.

Characteristics:

  • store multiple samples
  • handle both space and data
  • multiple labeled features (variables)
  • features can be scalars and vectors of different sizes
  • handle I/Os

Usage:

  • replaces classes Snapshot and Data
  • eliminates class Point
  • serve as a base class for class Space

Breaking change

Snapshot providers were re-implemented:

  • 3 providers file, function and job (This is a revolution !)
  • all providers are able to discover existing snapshots
  • all providers manage a cache of computed snapshots
  • all providers store all their cached snapshots in a single file ( #103 (closed) )
  • remove .wkdir directory used by job provider in snapshot directory tree
Edited by Cyril Fournier

Merge request reports