ENH: Potentially interesting packages

[BvB93]

Candidates

• NumExpr: Fast numerical array expression evaluator for Python, NumPy, PyTables, pandas, bcolz and more.
  • Support for >2D arrays appears to be somewhat limited; this could be problematic.
• Numba: Numba is an open source JIT compiler that translates a subset of Python and NumPy code into fast machine code.
  • Previous attempt at using Numba were rather unsuccessful, i.e. exceptions were appearing everywhere. Should revisit this at some point.
• Seaborn: Seaborn is a Python visualization library based on matplotlib. It provides a high-level interface for drawing attractive statistical graphics.
• yatiml: YAtiML is a small Python library that works with ruamel.yaml, adding functions for automatic type recognition to it. YAtiML is not a schema language like XSD and Relax-NG are for XML, or JSON Schema is for JSON. YAtiML is also not an Object/YAML mapper (YAML is already an object serialisation system, so you don't need an extra library for that). However, YAtiML does solve the same kind of problems, and more, so if you are looking for a schema language for YAML, YAtiML may actually be what you need.
  • Has some potentially interesting applications considering the extensive use of the .yaml format throughout Auto-FOX.

Implemented

• DASK: Dask is a flexible library for parallel computing in Python.
  • DASK is now used for accelerating the construction of angular distribution functions (ADFs).
• hdf5: Pythonic interface to the HDF5 binary data format.
  • The plans is to include this as a (non-optional) dependency for the next release for storing all Monte Carlo (MC) results. Will probably make it optional at a later point, the MC module reverting to the NPY format if HDF5 is not available.
  • hdf5 is now mandatory for use of the ARMC module(s).
  • hdf5 is now mandatory for all of Auto-FOX starting from 0.8.
• sphinx.ext.napoleon: Napoleon is a extension that enables Sphinx to parse both NumPy and Google style docstrings - the style recommended by Khan Academy.
  • The Google style looks very attractive; it has much less visual clutter compared to the standard docstring style.
  • The NumPy docstring style is now used in Auto-FOX.

Rejected

• pyparsing: The pyparsing module is an alternative approach to creating and executing simple grammars, vs. the traditional lex/yacc approach, or the use of regular expressions. The pyparsing module provides a library of classes that client code uses to construct the grammar directly in Python code.
  • Potentially usefull for parsing input files.
  • Rejected in favor of the (well-defined) .yaml input format.
• NumPy with OpenBlas support: A potentially easy way to improve NumPy performance (1). Precompiled versions are distributed on the conda-forge channel, which should make installation a bit easier.
  • Will probably include this as an optional dependency at some point, including (much needed) installation instructions.
  • This is considered the users responsibility.

[felipeZ]

You would probably like to add Numba to the list

[BvB93]

Hdf5 has been made a hard requirement for the Monte Carlo FF optimization (1).

[BvB93]

sphinx.ext.napoleon: NumPy docstring style implemented in #38.

[BvB93]

Plans for making hdf5 optional are on hold indefinitely.

[BvB93]

If installed, DASK is now used for accelerating the construction of angular distribution functions (ADFs).

[BvB93]

Added pyparsing.