Loosen spglib requirement slightly

docs/Troubleshooting.rst

@@ -41,46 +41,55 @@ error traceback.
 
 ``numpy`` Errors
 -------------------
-- A previous known issue with ``numpy``/``pymatgen`` is that it could give an error similar to this:
+A previous known issue with ``numpy``/``pymatgen`` is that it could give an error similar to this:
 
- .. code:: python
+.. code:: python
 
-  ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 88 from C header, got 80 from PyObject
+ ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 88 from C header, got 80 from PyObject
 
- This should be avoided with current versions of ``doped``, due to the package installation
- requirements (handled automatically by ``pip``), but depending on your ``python`` environment and
- previously-installed packages, it could possibly still arise. It occurs due to a recent change in the
- ``numpy`` C API in version ``1.20.0``, see
- `here <https://stackoverflow.com/questions/66060487/valueerror-numpy-ndarray-size-changed-may-indicate-binary-incompatibility-exp>`_
- for details.
- It should be fixed by reinstalling ``numpy`` and ``pymatgen`` (so that they play nice together), so
- that it is rebuilt with the new ``numpy`` C API:
+This should be avoided with current versions of ``doped``, due to the package installation
+requirements (handled automatically by ``pip``), but depending on your ``python`` environment and
+previously-installed packages, it could possibly still arise. It occurs due to a recent change in the
+``numpy`` C API in version ``1.20.0``, see
+`here <https://stackoverflow.com/questions/66060487/valueerror-numpy-ndarray-size-changed-may-indicate-binary-incompatibility-exp>`_
+for details.
+It should be fixed by reinstalling ``numpy`` and ``pymatgen`` (so that they play nice together), so
+that it is rebuilt with the new ``numpy`` C API:
 
- .. code:: bash
+.. code:: bash
 
-  pip install --force --no-cache-dir numpy==1.23
-  pip uninstall pymatgen
-  pip install pymatgen
+ pip install --force --no-cache-dir numpy==1.23
+ pip uninstall pymatgen
+ pip install pymatgen
 
 
 ``spglib`` Errors/Warnings
 --------------------------
-- A previous known issue with ``spglib`` is that it could give an error or warnings similar to:
+A previous known issue with ``spglib`` is that it could give an error or warnings similar to:
+
+.. code:: python
+
+ spglib: ssm_get_exact_positions failed (attempt=0).
+ spglib: No point group was found (line 405, ...).
+ ...
+ spglib: ssm_get_exact_positions failed (attempt=4).
+ spglib: get_bravais_exact_positions_and_lattice failed
+ spglib: ref_get_exact_structure_and_symmetry failed.
+
+This can be fixed by reinstalling ``spglib`` with ``conda install -c conda-forge spglib==2.0.2``.
+Sometimes installation with ``conda`` rather than ``pip`` is required, as ``conda`` will bundle the C
+and Fortran libraries, while using version ``2.0.2`` for now avoids some unnecessary warnings (see this
+`Issue <https://github.com/spglib/spglib/issues/338>`_ on the ``spglib`` GitHub for details).
 
- .. code:: python
+.. If this still does not remove these warnings, then setting the environment variable: ``SPGLIB_WARNING=OFF``
+and re-installing ``spglib`` with:
 
- spglib: ssm_get_exact_positions failed (attempt=0).
- spglib: No point group was found (line 405, ...).
- ...
- spglib: ssm_get_exact_positions failed (attempt=4).
- spglib: get_bravais_exact_positions_and_lattice failed
- spglib: ref_get_exact_structure_and_symmetry failed.
+.. .. code:: bash
 
- This can be fixed by reinstalling ``spglib`` with ``conda install -c conda-forge spglib==2.0.2``.
- Sometimes installation with ``conda`` rather than ``pip`` is required, as ``conda`` will bundle the C
- and Fortran libraries, while using version ``2.0.2`` for now avoids some unnecessary warnings (see this
- `Issue <https://github.com/spglib/spglib/issues/338>`_ on the ``spglib`` GitHub for details).
+.. pip uninstall spglib
+ pip install git+https://github.com/spglib/spglib --config-settings=cmake.define.SPGLIB_SHARED_LIBS=OFF
 
+.. should fix it. – STILL DOESN'T FULLY FIX IT YET (see doped_spglib_warnings.ipynb)
 
 ``ShakeNBreak``
 -------------------

pyproject.toml

@@ -22,7 +22,7 @@ dependencies = [
  "matplotlib>=3.5.2",
  "numpy>=1.21.0",
  "pymatgen>=2023.11.12", # limiting factor: POTCAR_STATS_PATH available from pymatgen.io.vasp.inputs
- "spglib==2.0.2", # dependent of pymatgen, but explicitly set to 2.0.2 here to avoid unnecessary warnings
+ "spglib<=2.0.2", # dependent of pymatgen, but explicitly set to <=2.0.2 here to avoid unnecessary warnings
  # see https://github.com/spglib/spglib/issues/338; remove this limit when issue resolved
  "pymatgen-analysis-defects>=2023.8.22",
  "shakenbreak>=3.3.1",
@@ -41,7 +41,10 @@ Package = "https://pypi.org/project/doped/"
 Repository = "https://github.com/SMTG-Bham/doped"
 
 [project.optional-dependencies]
-tests = ["pytest<8.2", "pytest-mpl>=0.16.1"]
+tests = [
+ "pytest<8.2", # with pytest==8.2.0, seemed to cause memory issue on GH Actions mpl tests, python=3.10
+ "pytest-mpl>=0.16.1"
+]
 docs = ["sphinx", "sphinx-book-theme", "sphinx_design"]
 analysis = [
  "easyunfold",