#13 scipy/numpy support

Core (6)
Jan Strube

I know that at the moment PAIDA is pure Python, but are
there plans to support numpy/scipy ?
The fitting would probably benefit a lot from it. At
the moment I'm using other options for fitting, because
it is just too slow. (Although the asymmetric fitter is
really nice and actually a feature that JAIDA doesn't have)

Matplotlib and PyTables support would probably also be
easier if that's supported.
Jython users probably already use the JAIDA
implementation anyways, which does have fitting code
written in java.


  • Koji Kishimoto

    Koji Kishimoto - 2006-08-23
    • labels: --> Core
    • milestone: --> Improvement
    • assigned_to: nobody --> korry
  • Koji Kishimoto

    Koji Kishimoto - 2006-08-23

    Logged In: YES

    Both numpy/scipy and PAIDA have almost the same fitting logics and they are
    pure Python codes. Then, I've tested the fitting performance of numpy/scipy
    array as a histogram data storage and gotten a conclusion that the current
    data array implementation is equal speed or faster than them.
    But, because scipy has various useful functions, scipy support is still an
    attractive idea worth considering.
    As far as the fitting speed, using already-existing C/C++ codes as an
    _optional_ fitting engine would be also a good idea!

  • Jan Strube

    Jan Strube - 2006-10-09

    Logged In: YES

    That wouold be great.
    I noticed that there are some subtleties involved in
    histogramming numpy arrays, as they can contain nan (not a
    number) values.


Log in to post a comment.