Skip to content

Releases: dsavoiu/kafe

kafe release 1.3.3

26 Feb 15:07
Compare
Choose a tag to compare

This is a maintenance release that addresses the incompatibility between kafe and iminuit 2.0 or later.
Running with recent iminuit versions now prompts users to downgrade to 1.5.4.

kafe release 1.3.2

07 Oct 12:33
Compare
Choose a tag to compare

This release fixes a number of small issues, mainly related to changes/deprecations in the matplotlib API.

Changes

  • FIX: ensure backend fallback mechanism remains active
    despite explicit call to matplotlib.use, which otherwise leads
    to failures if the backend is not available (or when running in
    headless mode)
  • FIX: move imports of matplotlib.pyplot to plot methods to avoid
    importing them prematurely
  • FIX: handle deprecation of axis suffixes for keyword arguments
    passed to constructors of axis scale objects starting with
    matplotlib 3.3.0 (e.g. now specify base instead of
    basex/basey for LogScale)
  • FIX: replace call to inverse_transformed completely
    by equivalent transformed(transform.inverted())
  • FIX: use Legend.set_draggable(True) and fall back to deprecated
    Legend.draggable() on failure for backwards compatibility.

Bugfix release

01 Aug 14:10
Compare
Choose a tag to compare

Some output files were empty when running under Python 3. This version provides a hotfix
by forcing line-buffering on output streams and adding flush() and close() commands
where appropriate.

kafe release 1.3.0

14 Jun 17:58
Compare
Choose a tag to compare

New features

  • now runs with both Python 2 and 3 (many thanks to @smartsammler)
  • can do multi-model fits with shared parameters (module Multifit by @JoergSchindler)

Fixes and improvements

Some bugs were fixed and compatibility with other tools has been improved:

  • kafe now works in Jupyter Notebooks

  • fix compatibility issue with matplotlib 2.0

  • revise minimum recommended versions for scipy, numpy and iminuit modules

  • drop requirement for Qt/PyQt due to difficulties in getting it to work with conda

    • Tk is now the default backend for matplotlib
  • better output stream management (fails gracefully when redirection is unsupported, instead of crashing)