add h5py to setup.py's extra_requires (#3711)

* add h5py to setup.py's extra_requires
  - fix #3701
  - new 'extra_coordinates' target in extra_requires where we can
    add pip-installable packages for additional formats that are not
    installed by default
  - keep minimum version of h5py at >= 2.1.0
* minimum h5py >= 2.10
  - GitHub actions
  - maintainer/conda/environment.yml for readthedocs
  - Azure pipelines
* update extra_deps in setup.py
  - include chemfiles>=0.10 in extra_formats
  - added comment to remove AMBER in 3.0.0
* updated CHANGELOG
* updated sklearn --> scikit-learn in setup.py
* deprecate extras_require "AMBER" target for 2.4

Co-authored-by: Irfan Alibay <[email protected]>

.github/actions/setup-deps/action.yaml

@@ -54,7 +54,7 @@ inputs:
   clustalw:
     default: 'clustalw=2.1'
   h5py:
-    default: 'h5py'
+    default: 'h5py>=2.10'
   joblib:
     default: 'joblib>=0.12'
   netcdf4:

azure-pipelines.yml

@@ -92,7 +92,7 @@ jobs:
       pytest-xdist
       scikit-learn
       scipy
-      h5py
+      h5py>=2.10
       tqdm
       threadpoolctl
       fasteners

maintainer/conda/environment.yml

@@ -9,7 +9,7 @@ dependencies:
   - fasteners
   - griddataformats
   - gsd
-  - h5py
+  - h5py>=2.10
   - hypothesis
   - joblib>=0.12
   - matplotlib==3.2.2

package/CHANGELOG

@@ -29,12 +29,17 @@ Enhancements
     (CZI Performance track, PR #3683)
 
 Changes
+  * To add additional optional packages for different file formats when
+    using `pip` use the new "extra_formats" extras_require target as in
+    `pip install ./package[extra_formats]` (Issue #3701, PR #3711)
   * Minimum supported NumPy version is now 1.20 (1.21 for macosx-arm64)
     as per NEP29 (PR #3737)
   * Narrowed variable scope to reduce use of OpenMP `private` clause (PR #3706, PR
     #3728)
 
 Deprecations
+  * The extras_requires target "AMBER" for `pip install ./package[AMBER]`
+    will be removed in 2.4. Use "extra_formats". (Issue #3701, PR #3711)
   * Deprecate `density` parameter in favor of `norm` in InterRDF_s
     (Issue #3687)
 

package/setup.py

@@ -63,15 +63,8 @@
     print('Please upgrade your version of Python.')
     sys.exit(-1)
 
-if sys.version_info[0] < 3:
-    import ConfigParser as configparser
-else:
-    import configparser
-
-if sys.version_info[0] >= 3:
-    from subprocess import getoutput
-else:
-    from commands import getoutput
+import configparser
+from subprocess import getoutput
 
 # NOTE: keep in sync with MDAnalysis.__version__ in version.py
 RELEASE = "2.3.0-dev0"
@@ -654,14 +647,19 @@ def long_description(readme):
           # you might prefer to use the version available through your
           # packaging system
           extras_require={
-              'AMBER': [
+              'AMBER': [           # REMOVE for 2.4.0, use 'extra_formats'
                   'netCDF4>=1.0',  # for fast AMBER writing, also needs HDF5
               ],
+              'extra_formats': [   # additional file formats
+                  'netCDF4>=1.0',  # for fast AMBER writing, also needs HDF5
+                  'h5py>=2.10',    # H5MD
+                  'chemfiles>=0.10',  # multiple formats supported by chemfiles
+                  ],
               'analysis': [
                   'seaborn',  # for annotated heat map and nearest neighbor
                               # plotting in PSA
-                  'sklearn',  # For clustering and dimensionality reduction
-                              # functionality in encore
+                  'scikit-learn',  # For clustering and dimensionality
+                                   # reduction functionality in encore
                   'tidynamics>=1.0.0', # For MSD analysis method
               ],
           },