Merge pull request #925 from hackingmaterials/ml-evs/deps-tinkering

Reintroduce pandas pin and update tests

.github/ISSUE_TEMPLATE.md

@@ -1,6 +1,8 @@
 Please submit help issues to:
-https://discuss.matsci.org/c/matminer/automatminer
+https://matsci.org/c/matminer/16
 
 The Github issues is no longer used except for internal development purposes.
 
 If you are unable to use the Discourse forum, you may submit an issue here, but you must **clearly** state the reason you are unable to use the Discourse forum in your ticket. Otherwise, your issue will be **closed** without response.
+
+Please search both the forums and former GitHub issues before raising a ticket to avoid duplicating discussions.

.github/workflows/test.yml

@@ -3,6 +3,9 @@
 # The ext package is also only tested in this workflow. Coverage is also computed based on this platform.
 name: Testing
 
+env:
+  PYTEST_ADDOPTS: "--color=yes"
+
 on:
   push:
     branches:

README.md

@@ -7,7 +7,7 @@ matminer is a library for performing data mining in the field of materials scien
 - **[Help/Support Forum](https://matsci.org/c/matminer/16)**
 - **[Source Repository](https://github.com/hackingmaterials/matminer)**
 
-matminer supports Python 3.8+.
+matminer supports Python 3.9+.
 
 #### Related packages:
 

docs_rst/installation.rst

@@ -7,7 +7,9 @@
 Installing matminer
 ===================
 
-Matminer requires Python 3.6+.
+Matminer requires Python 3.9+.
+
+It is strongly recommend that you install it in a virtual environment (using e.g., conda, Python's venv, or related tools) to avoid conflicts with other packages.
 
 There are a couple of quick and easy ways to install matminer (see also some **tips** below):
 
@@ -28,6 +30,9 @@ or, to install matminer in your user home folder, run the command:
 
 To update matminer, simply type ``pip install --upgrade matminer``.
 
+This will install the latest version of all of matminer's dependencies.
+In cases of future incompatibility, you can install the specific versions used for testing matminer by following the development instructions below.
+
 Install in development mode
 -----------------------------
 
@@ -42,15 +47,16 @@ and then enter the cloned repository/folder to install in developer mode:
 .. code-block:: bash
 
     $ cd matminer
-    $ python setup.py develop
+    $ pip install -r requirements/requirements-ubuntu_py3.11.txt  # or corresponding file for your platform
+    $ pip install -e .
 
-To update matminer, enter your cloned folder and type ``git pull`` followed by ``python setup.py develop``.
+To update matminer, simply enter your cloned folder and type ``git pull``g
 
 
 Tips
 ----
 
-* Make sure you are using Python 3.6 or higher
+* Make sure you are using a compatible Python version (3.9+ at time of writing).
 * If you have trouble with the installation of a component library (sympy, pymatgen, mdf-forge, etc.), you can try to run ``pip install <<component>>`` or (if you are using `Anaconda <https://www.anaconda.com/distribution/>`_) ``conda install <<component>>`` first, and then re-try the installation.
 
     - For example, installing pymatgen on a Windows platform is easiest with Anaconda via ``conda install -c conda-forge pymatgen``.

matminer/featurizers/composition/tests/test_alloy.py

@@ -160,7 +160,7 @@ def test_WenAlloys(self):
             "Shear modulus mean": 79.48600137832061,
             "Shear modulus strength model": -0.009636621848440554,
             "Total weight": 57.243028243301005,
-            "VEC mean": 8.447250438374748,
+            "VEC mean": 8.395723406331793,
             "Weight Fraction": "Fe0.6048579375087819 C0.00019995792415715736 "
             "Mn0.0005000210911858884 Si0.0005004488909678273 "
             "Cr9.991733798026916e-05 Ni0.19686472127404955 "

matminer/featurizers/site/external.py

@@ -123,7 +123,7 @@ def from_preset(cls, preset):
         """
         valid_presets = ["formation_energy"]
         if preset == "formation_energy":
-            return cls(6, 8, 8, 0.4, True, "gto", True)
+            return cls(rcut=6, nmax=8, lmax=8, sigma=0.4, periodic=True, rbf="gto", crossover=True, compression=False)
         else:
             raise ValueError(f"'{preset}' is not a valid preset. Choose from {valid_presets}")
 

matminer/featurizers/site/rdf.py

@@ -428,7 +428,7 @@ def featurize(self, struct, idx):
         # Generate cos(theta) between neighbor pairs for each central site.
         # Also, retain data on neighbor distances for each pair
         # process with matrix algebra, we really need the speed here
-        data = np.array(list(neighbor_tuples))
+        data = np.array(list(neighbor_tuples), dtype=object)
         v1, v2 = np.vstack(data[:, 0, 0]), np.vstack(data[:, 1, 0])
         distances = data[:, :, 1]
         neighbor_pairs = np.concatenate(

matminer/featurizers/site/tests/test_rdf.py

@@ -86,7 +86,6 @@ def test_grdf(self):
             [f"Gaussian center={i} width=1.0" for i in np.arange(10.0)],
         )
 
-    @unittest.skip("Test does not pass with latest deps")
     def test_afs(self):
         f1 = Gaussian(1, 0)
         f2 = Gaussian(1, 1)

matminer/featurizers/structure/bonding.py

@@ -533,7 +533,7 @@ def fit(self, X, y=None):
         """
         unpadded_bobs = [self.bag(s, return_baglens=True) for s in X]
         bonds = [list(bob.keys()) for bob in unpadded_bobs]
-        bonds = np.unique(sum(bonds, []))
+        bonds = np.unique(np.array(sum(bonds, []), dtype=object))
         baglens = [0] * len(bonds)
 
         for i, bond in enumerate(bonds):

matminer/featurizers/structure/tests/test_bonding.py

@@ -60,7 +60,6 @@ def test_bondfractions(self):
         np.testing.assert_array_equal(df["Al - Al bond frac."].to_numpy(), [0.0, 0.0])
         np.testing.assert_array_equal(df["Ni - Ni bond frac."].to_numpy(), [0.0, 0.5])
 
-    @unittest.skip("Test does not pass with latest deps")
     def test_bob(self):
         # Test a single fit and featurization
         scm = SineCoulombMatrix(flatten=False)

matminer/featurizers/structure/tests/test_sites.py

@@ -11,7 +11,6 @@
 
 
 class StructureSitesFeaturesTest(StructureFeaturesTest):
-    @unittest.skip("Test does not pass with latest deps")
     def test_sitestatsfingerprint(self):
         # Test matrix.
         op_struct_fp = SiteStatsFingerprint.from_preset("OPSiteFingerprint", stats=None)

matminer/featurizers/utils/grdf.py

@@ -88,7 +88,6 @@ def __call__(self, r_ij):
         return np.logical_and(
             np.greater_equal(r_ij, self.start),
             np.less(r_ij, self.start + self.width),
-            dtype=np.float64,
         )
 
     def volume(self, cutoff):

matminer/featurizers/utils/tests/test_grdf.py

@@ -32,7 +32,6 @@ def test_gaussian(self):
         self.assertIn("width=4", name)
         self.assertIn("center=4", name)
 
-    @unittest.skip("Test does not pass with latest deps")
     def test_histogram(self):
         h = Histogram(1, 4)
         np.testing.assert_array_almost_equal([0, 1, 0], h([0.5, 2, 5]))

setup.py

@@ -54,16 +54,15 @@ def local_version(version):
         include_package_data=True,
         zip_safe=False,
         install_requires=[
-            "numpy>=1.20.1,<2",
-            "requests",
-            "pandas",
-            "tqdm",
-            "pymongo",
-            "future",
-            "scikit_learn",
-            "sympy",
-            "monty",
-            "pymatgen",
+            "numpy >= 1.23, < 2",
+            "requests ~= 2.31",
+            "pandas ~= 1.5",
+            "tqdm ~= 4.66",
+            "pymongo ~= 4.5",
+            "scikit_learn ~= 1.3",
+            "sympy ~= 1.11",
+            "monty >= 2023",
+            "pymatgen >= 2023",
         ],
         extras_require=extras_require,
         classifiers=[