Import bug fixes, readme + action updates, and addressed divion error…

… in FS step

.github/workflows/python-package.yml

@@ -9,7 +9,8 @@ name: Python package
 on:
   push:
     branches:
-      -stable
+      - stable
+      - main
   workflow_dispatch:
 
 jobs:
@@ -50,7 +51,11 @@ jobs:
         path: ./wheelhouse/*.whl
 
   release:
-  # https://packaging.python.org/en/latest/guides/publishing-package-distribution-releases-using-github-actions-ci-cd-workflows/
+    # https://packaging.python.org/en/latest/guides/publishing-package-distribution-releases-using-github-actions-ci-cd-workflows/
+
+    # https://docs.github.com/en/actions/learn-github-actions/variables#default-environment-variables
+    # Only create release if pull-target was 'stable' or work-flow was triggered on 'stable'
+    if: github.ref_name == 'stable' || github.base_ref == 'stable'
     needs:
       - build
     runs-on: ubuntu-latest
@@ -59,6 +64,9 @@ jobs:
       contents: write  # Necessary for creating releases apparently.
 
     steps:
+
+      - name: Echo Branch
+        run: echo "${{ github.ref_name }} ${{ github.base_ref }}"
 
       - name: Download Artifacts
         uses: actions/download-artifact@v3

README.md

@@ -3,15 +3,26 @@
 ## Description
 ``mssm`` is a toolbox to estimate Generalized Additive Mixed Models (GAMMs) semi Markov-switching GAMMs (sMs-GAMMs) and sMs Impulse Response GAMMs (sMs-IR-GAMMs).
 The ``main`` branch is updated frequently to reflect new developments. The ``stable`` branch should reflect the latest releases. if you don't need the latest functionality, you
-should install from the ``stable`` branch.
+should install from the ``stable`` branch (see below for instructions).
 
 ## Installation
-There are two options!
 
-Option 1 is to build directly from source. This requires ``conda`` or an installation of [eigen](https://eigen.tuxfamily.org/index.php?title=Main_Page)(``setup.py`` then expects ``eigen`` in "usr/include/eigen3"). If you have ``conda`` installed
-[install eigen from conda-forge](https://anaconda.org/conda-forge/eigen). After cloning and navigating into the downloaded repository you can then install via ``pip install . ``.
+The easiest option is to install from pypi via ``pip``.
+
+1) Setup a conda environment with python > 3.10
+2) Install mssm via ``pip``
+
+Currently, mssm can only be installed from [test.pypi](https://test.pypi.org/project/mssm/#description) because sign-ups are disabled for pypi. So to install ``mssm`` run:
 
-Option 2 (currently being worked on) is to install one of the releases from the ``stable`` branch.
+```
+conda create -n mssm_env python=3.10
+pip install -i https://test.pypi.org/simple/ mssm
+```
+
+Once ``mssm`` can be published to pypi, you can replace the second line with ``pip install mssm``. Note: pypi will only reflect releases on the ``stable`` branch. Pushes to main will however continue to be distributed to test.pypi, so if you need the latest changes you can get them from there.
+
+You can also build directly from source. This requires ``conda`` or an installation of [eigen](https://eigen.tuxfamily.org/index.php?title=Main_Page)(``setup.py`` then expects ``eigen`` in "usr/local/include/eigen3". This will probably not work on windows.). Once you have ``conda`` installed,
+[install eigen from conda-forge](https://anaconda.org/conda-forge/eigen). After cloning and navigating into the downloaded repository you can then install via ``pip install . ``.
 
 ## To get started
 

pyproject.toml

@@ -11,7 +11,7 @@ dependencies=["numpy >= 1.24.1",
               "pandas >= 1.5.3",
               "scipy >= 1.10.0"]
 name = "mssm"
-version = "0.1.0"
+version = "0.1.1"
 authors = [
   { name="Joshua Krause", email="jokra001@proton.me" }
 ]

src/mssm/src/python/gamm_solvers.py

@@ -3,7 +3,7 @@
 import warnings
 from .exp_fam import Family,Gaussian
 from .penalties import PenType,id_dist_pen,translate_sparse
-from ..cpp import cpp_solvers
+import cpp_solvers
 
 def cpp_chol(A):
    return cpp_solvers.chol(A)
@@ -23,6 +23,12 @@ def step_fellner_schall_sparse(gInv,emb_SJ,Bps,cCoef,cLam,scale,verbose=False):
 
   num = max(0,(gInv @ emb_SJ).trace() - Bps)
   denom = max(0,cCoef.T @ emb_SJ @ cCoef)
+
+  # Especially when using Null-penalties denom can realisitically become
+  # equal to zero: every coefficient of a term is penalized away. In that
+  # case num /denom is not defined so we return directly.
+  if denom <= 0: # Prevent overflow
+     return 1e+7
   nLam = scale * (num / denom) * cLam
   nLam = max(nLam,1e-7) # Prevent Lambda going to zero
   nLam = min(nLam,1e+7) # Prevent overflow