Hi Folks,

Larch 0.9.43 is now available, with installers for Windows, MacOS, and Linux at
   https://millenia.cars.aps.anl.gov/xraylarch/installation.html

If you already have installed Larch, you should be able to update to the latest version with
    ~>  conda update -c gsecars xraylarch

from a terminal on Linux or Mac OSX.  On Windows, you may have to specify the full path with something like:
   C:\Users\<YourName>\AppData\Local\Continuum\xraylarch\Scripts\conda.exe -c gsecars xraylarch

If you have any trouble upgrading, you can simply remove the xraylarch installation folder and reinstall. If you would like to install Larch into a different Python environment, please read  https://xraypy.github.io/xraylarch/installation.html

Version 0.9.43 has several improvements to the XAS Viewer GUI application, including:
  -  better handling of read/write cycles of multiple Athena project files.
  -  better default normalization and better control over normalization, especially for XANES data.
  -  improved PCA analysis (in part from the recent discussion here with Joselaine Cáceres Gonzalez),  so that it now reports IND values to help determine the number of components and reports eigenvalues from simple matrix inversion (not SVD) so that they more closely match those in the XAFS literature.
  -  addition of Partial Least Squares and LASSO regression analysis for selection and prediction of external quantities (valence, for example) based on training sets of XANES data with known values for these quantities. This machine-learning approach is based on work of M Dyar, et al in a series of papers over the past several years.


I encourage and request anyone interested, and especially MacOS users, to try out the XAS Viewer app and let us know what needs improvement.

For people interested in using Larch from Python, Version 0.9.43 includes a complete refactoring of the code to make Larch work better as a "normal" Python library.  Specifically, the previous code organization with most of the real analysis code in "plugins" has been replaced with all code now placed within the main larch module, still organized by topic.  This improves packaging -- `pip install xraylarch` can now work.  This code reorganization also means that the python programmer can use more normal imports to get at the Larch library so that one can simply do


    >>> from larch.xafs import pre_edge, autobk, xftf


We also fixed a serious and deep flaw that selected a matplotlib plotting library too early, making larch difficult to use with Jupyter, Spyder, or other Qt-based GUIs.  This is now fixed, and import statements like the one above can be seamlessly used in Jupyter notebooks.

I should note that the documentation and examples are definitely lagging behind the code especially regarding the most recent developments, but this will be worked on.  If you having any questions, trouble, or suggestions on any part of Larch, please let us know.

--Matt Newville