Advanced Numpy

• Abstract
• Advanced Numpy
• Life of ndarray
  • It’s...
  • Block of memory
  • Data types
    • The descriptor
    • Example: reading .wav files
    • Casting and re-interpretation/views
  • Indexing scheme: strides
    • Main point
    • Example: fake dimensions with strides
    • Broadcasting
    • More tricks: diagonals
    • CPU cache effects
    • Example: inplace operations (caveat emptor)
  • Findings in dissection
• Universal functions
  • What they are?
  • Exercise: building an ufunc from scratch
  • Solution: building an ufunc from scratch
  • Generalized ufuncs
• Interoperability features
  • Sharing multidimensional, typed data
  • The old buffer protocol
  • The old buffer protocol
  • Array interface protocol
  • The new buffer protocol: PEP 3118
  • PEP 3118 – details
• Siblings: chararray, maskedarray, matrix
• Summary
• Hit list of the future for Numpy core
• Contributing to Numpy/Scipy
  • Why
  • Reporting bugs
  • Contributing to documentation
  • Contributing features
  • How to help, in general
• Python 2 and 3, single code base
  • The case of Numpy
  • The case of Scipy
  • How?