release date: Q1 2020

Expected:

• Numpy-1.18, Matplotlib-3.2.0, Pandas-1.0, Bokeh-2.0, Jupyterlab-2.0
• maybe spyder-4.1.0
• better Jupyter compatible Python-vscode (good compatibility delayed to late april 2020 https://github.com/microsoft/vscode-python/issues/3429)
• a minimal "dot" version of Python-3.9

Cleanup:

• remove ipythreejs/ipydatawidgets : when not supported, trashes jupyterlab build
• maybe a cleanup based on factual numbers https://pyviz.org/tools.html

Focus of the release:

• vscode refined integration : (improving but still not good in this release)
  • : february update (less complete than expected)
  • : mid-march update (still won't include ipywidgets)
• cleanup & simplify further the build steps:
  • JupyterLab-2 (future help)

In a future release:

• Tensorflow-cpu-2.2.0rc1 for Python-3.8 was released march 19th, but not for windows, ... so for next Winpython release,
• once again screwed up in packages.ini (should be automated or code more resilient to error)
• explore again distance to build in the cloud.
• reduce need for an initial full WinPython environnment.
• Spyder-5 compatible with PySide2 ... Q4 2020 ? (when Anaconda will switch)
• BLIS / OpenBLAS: looking after MKL to benefit of AMD Zen2+ https://twitter.com/ogrisel/status/1229457423443939328

Hopes/Wishes/Strategy for coming 2020:

• Python-3.9 simplifies Asyncio (inspired per Trio) and multi-core (sub-interpreters) Q4 2020
• Python GIL-less future is emerging : https://morepypy.blogspot.com/2019/12/hpy-kick-off-sprint-report.html
• python-vscode becomes DataScientist super-friendly (Q2 2020 ?)
• Python hits 31% on PYPL. (will equal Java record of 30.9% in December 2007)