Adding version information to kwf files (#212)

Currently `.kwf` files can be quite fickle because the format relies on the structure of the `Workflow` object itself, so whenever we change the `Workflow` class we lose the ability to read old `kwf` files.

Now, we have added explicitly the version number to the `Workflow` class. This will warn users if they are loading a `.kwf` file that was generated with a different version of `koopmans` than that which generated it, and, if the reading fails, they can always revert to the version of `koopmans` that generated the file in the first place.

.github/workflows/docs.yml

@@ -11,7 +11,7 @@ jobs:
   docs:
     runs-on: ubuntu-latest
     steps:
-    - uses: actions/checkout@v2
+    - uses: actions/checkout@v3
     - uses: ammaraskar/sphinx-action@master
       with:
         docs-folder: "docs/"

.github/workflows/tests.yml

@@ -18,11 +18,11 @@ jobs:
             os: [ubuntu-latest, macOS-latest]
       steps:
          - name: Checkout
-           uses: actions/checkout@v2
+           uses: actions/checkout@v3
            with:
               submodules: true
          - name: Set up Python 3.8
-           uses: actions/setup-python@v2
+           uses: actions/setup-python@v4
            with:
               python-version: 3.8
          - name: Install dependencies

.github/workflows/tutorials.yml

@@ -18,11 +18,11 @@ jobs:
             os: [ubuntu-latest, macOS-latest]
       steps:
          - name: Checkout
-           uses: actions/checkout@v2
+           uses: actions/checkout@v3
            with:
               submodules: true
          - name: Set up Python 3.8
-           uses: actions/setup-python@v2
+           uses: actions/setup-python@v4
            with:
               python-version: 3.8
          - name: Install dependencies

.github/workflows/typechecking.yml

@@ -16,11 +16,11 @@ jobs:
          max-parallel: 3
       steps:
          - name: Checkout
-           uses: actions/checkout@v2
+           uses: actions/checkout@v3
            with:
               submodules: false
          - name: Set up Python 3.8
-           uses: actions/setup-python@v2
+           uses: actions/setup-python@v4
            with:
               python-version: 3.8
          - name: Install dependencies

.github/workflows/update_citation.yml

@@ -13,12 +13,12 @@ jobs:
       runs-on: ubuntu-latest
       steps:
          - name: Checkout
-           uses: actions/checkout@v2
+           uses: actions/checkout@v3
          - name: Get branch name
            id: branch-name
            uses: tj-actions/branch-names@v6
          - name: Set up Python 3.8
-           uses: actions/setup-python@v2
+           uses: actions/setup-python@v4
            with:
               python-version: 3.8
          - name: Install dependencies

src/koopmans/__init__.py

@@ -7,4 +7,4 @@
 else:
     import importlib_metadata as metadata
 
-__version__ = metadata.version('koopmans')
+__version__: str = metadata.version('koopmans')

src/koopmans/workflows/_ml.py

@@ -1,13 +1,11 @@
 """
 Written by Yannick Schubert Jul 2022
 """
-import copy
 import os
 from pathlib import Path
-from typing import Any, Dict, Tuple, Optional, List
+from typing import Any, Dict, List, Tuple
 
 import numpy as np
-from deepdiff import DeepDiff
 from sklearn.metrics import mean_absolute_error as mae
 
 from koopmans import calculators, ml, utils
@@ -75,7 +73,7 @@ def _run(self):
 
         # Specify for which bands we want to compute the decomposition
         self.bands_to_extract = self.bands.to_solve
-        self.num_bands_occ = [len([band for band in self.bands if (band.filled == True and band.spin == spin)])
+        self.num_bands_occ = [len([band for band in self.bands if (band.filled and band.spin == spin)])
                               for spin in [0, 1]]
         self.num_bands_to_extract = [len([band for band in self.bands_to_extract if band.filled == filled])
                                      for filled in [True, False]]
@@ -96,7 +94,8 @@ def _run(self):
                 self.extract_input_vector_from_self_hartrees()
             else:
                 raise ValueError(
-                    f"{self.ml.input_data_for_ml_model} is currently not implemented as a valid input for the ml model.")
+                    f"{self.ml.input_data_for_ml_model} is currently not implemented as a valid input for the ml "
+                    "model.")
 
     def extract_input_vector_from_self_hartrees(self):
         """
@@ -109,7 +108,7 @@ def extract_input_vector_from_self_hartrees(self):
             else:
                 filled_str = 'emp'
             np.savetxt(self.dirs['SH'] / f"SH.orbital.{filled_str}.{band.index}.txt",
-                       np.array([SH[band.spin][band.index-1]]))
+                       np.array([SH[band.spin][band.index - 1]]))
             self.input_vectors_for_ml[f"SH.orbital.{filled_str}.{band.index}"] = SH
 
         return
@@ -158,11 +157,11 @@ def convert_bin2xml(self):
 
     def check_if_bin2xml_is_complete(self) -> bool:
         """
-        Checks if the bin2xml conversion was already performed. 
-        
-        For this we first check if the next step is complete, i.e. the decomposition into spherical harmonics and 
-        radial basis functions. If we haven't yet computed the decomposition, we check if this step, i.e. the 
-        conversion to xml, was already performed. This way we can delete the (possibly large) binary and xml files 
+        Checks if the bin2xml conversion was already performed.
+
+        For this we first check if the next step is complete, i.e. the decomposition into spherical harmonics and
+        radial basis functions. If we haven't yet computed the decomposition, we check if this step, i.e. the
+        conversion to xml, was already performed. This way we can delete the (possibly large) binary and xml files
         and only keep the expansion coefficients and still be able to restart the calculation.
         """
 
@@ -373,13 +372,10 @@ def get_alpha_from_file_for_debugging(self, band: Band) -> Tuple[float, float]:
         if self.parameters.spin_polarized:
             raise NotImplementedError('Need to check implementation')
         assert isinstance(band.index, int)
-        return alphas[0][band.index-1], 0.0
+        return alphas[0][band.index - 1], 0.0
 
     @classmethod
     def fromdict(cls, dct: Dict[str, Any], **kwargs) -> Workflow:
         calc_that_produced_orbital_densities = dct.pop('calc_that_produced_orbital_densities')
         return super(MLFittingWorkflow, cls).fromdict(
             dct, calc_that_produced_orbital_densities=calc_that_produced_orbital_densities, **kwargs)
-
-    def __eq__(self, other):
-        return DeepDiff(self, other) == {}

src/koopmans/workflows/_workflow.py

@@ -102,6 +102,7 @@ class Workflow(ABC):
     pseudo_dir: Path
     projections: ProjectionBlocks
     parent: Optional[Workflow]
+    version: str
 
     def __init__(self, atoms: Atoms,
                  pseudopotentials: Dict[str, str] = {},
@@ -114,6 +115,7 @@ def __init__(self, atoms: Atoms,
                  plotting: Union[Dict[str, Any], settings.PlotSettingsDict] = {},
                  ml: Union[Dict[str, Any], settings.MLSettingsDict] = {},
                  autogenerate_settings: bool = True,
+                 version: Optional[str] = None,
                  **kwargs: Dict[str, Any]):
 
         # Parsing parameters
@@ -308,7 +310,8 @@ def __init__(self, atoms: Atoms,
                     calc_params[key] = value
                     match = True
             # if not a calculator, workflow, or plotting keyword, raise an error
-            if not match and not self.parameters.is_valid(key) and not self.plotting.is_valid(key) and not self.ml.is_valid(key):
+            if not match and not self.parameters.is_valid(key) and not self.plotting.is_valid(key) \
+                    and not self.ml.is_valid(key):
                 raise ValueError(f'{key} is not a valid setting')
 
         # Adding excluded_bands info to self.projections
@@ -319,6 +322,13 @@ def __init__(self, atoms: Atoms,
                     label += f'_{spin}'
                 self.projections.exclude_bands[spin] = self.calculator_parameters[label].get('exclude_bands', [])
 
+        # Version number (important when loading workflows from .kwf files)
+        from koopmans import __version__
+        self.version = version if version is not None else __version__
+        if self.version != __version__:
+            utils.warn(f'You are using version {__version__} of koopmans, but this workflow was generated with '
+                       f'version {self.version}. Proceed with caution.')
+
     def __eq__(self, other: Any):
         if isinstance(other, Workflow):
             return self.__dict__ == other.__dict__
@@ -517,7 +527,8 @@ def _run_sanity_checks(self):
 
         # Make sanity checks for the ML model
         if self.ml.use_ml:
-            utils.warn("Predicting screening parameters with machine-learning is an experimental feature; proceed with caution")
+            utils.warn("Predicting screening parameters with machine-learning is an experimental feature; proceed with "
+                       "caution")
             if self.parameters.task not in ['trajectory', 'convergence_ml']:
                 raise NotImplementedError(
                     f'Using the ML-prediction for the {self.parameter.task}-task has not yet been implemented.')
@@ -526,17 +537,20 @@ def _run_sanity_checks(self):
                     f"Using the ML-prediction for the {self.parameters.method}-method has not yet been implemented")
             if self.parameters.functional != 'ki':
                 raise NotImplementedError(
-                    f'Using the ML-prediction for the {self.parameters.functional}-functional has not yet been implemented.')
+                    f'Using the ML-prediction for the {self.parameters.functional}-functional has not yet been '
+                    'implemented.')
             if self.parameters.init_orbitals != 'mlwfs':
                 raise NotImplementedError(
-                    f'Using the ML-prediction for {self.parameters.init_orbitals}-init orbitals has not yet been implemented.')
+                    f'Using the ML-prediction for {self.parameters.init_orbitals}-init orbitals has not yet been '
+                    'implemented.')
             if self.parameters.init_empty_orbitals != self.parameters.init_orbitals:
                 raise NotImplementedError(
-                    f'Using the ML-prediction for using different init orbitals for empty states than for occupied states has not yet been implemented.')
+                    'Using the ML-prediction for using different init orbitals for empty states than for occupied '
+                    'states has not yet been implemented.')
             if self.parameters.spin_polarized:
-                utils.warn(f'Using the ML-prediction for spin-polarised systems has not yet been extensively tested.')
+                utils.warn('Using the ML-prediction for spin-polarised systems has not yet been extensively tested.')
             if not all(self.atoms.pbc):
-                utils.warn(f'Using the ML-prediction for non-periodic systems has not yet been extensively tested.')
+                utils.warn('Using the ML-prediction for non-periodic systems has not yet been extensively tested.')
             if self.parameters.orbital_groups:
                 utils.warn('Using orbital_groups has not yet been extensively tested.')
             if not np.all(self.atoms.cell.angles() == 90.0):
@@ -564,7 +578,8 @@ def convert_to_list(param, type):
             r_mins = convert_to_list(self.ml.r_min, float)
             r_maxs = convert_to_list(self.ml.r_max, float)
 
-            # check that each n_max, l_max, r_max and r_min are greater or equal to 0 and that r_min is smaller than r_max
+            # check that each n_max, l_max, r_max and r_min are greater or equal to 0 and that r_min is smaller than
+            # r_max
             for n_max in n_maxs:
                 if not n_max > 0:
                     raise ValueError(f"n_max has to be larger than zero. The provided value is n_max={n_max}")
@@ -576,7 +591,8 @@ def convert_to_list(param, type):
                     raise ValueError(f"r_min has to be equal or larger than zero. The provided value is r_min={r_min}")
                 if r_min < 0.5:
                     utils.warn(
-                        f"Small values of r_min (<0.5) can lead to problems in the construction of the radial basis. The provided value is r_min={r_min}.")
+                        "Small values of r_min (<0.5) can lead to problems in the construction of the radial basis. "
+                        f"The provided value is r_min={r_min}.")
             for r_max in r_maxs:
                 if not any(r_min < r_max for r_min in r_mins):
                     raise ValueError(f"All provided values of r_min are larger or equal to r_max={r_max}.")
@@ -940,6 +956,12 @@ def todict(self):
 
     @classmethod
     def fromdict(cls, dct: Dict[str, Any], **kwargs) -> Workflow:
+
+        # Remove __koopmans_name/module__ if present (won't happen if the encoder was used, but will happen if
+        # todict and fromdict are used directly)
+        dct.pop('__koopmans_name__', None)
+        dct.pop('__koopmans_module__', None)
+
         wf = cls(atoms=dct.pop('atoms'),
                  parameters=dct.pop('parameters'),
                  calculator_parameters=dct.pop('calculator_parameters'),