Use attrs for IOData class

Fixes theochem#73

Not all pylint exclusions can be removed, essentially due to pylint
bugs resulting in many false negatives, see

https://stackoverflow.com/questions/47972143/using-attr-with-pylint
pylint-dev/pylint#1694

The good solution is to disable all type-checking warnings of
pylint and to use a proper type checker like mypy instead.

Related changes:

- reaction_coordinate, ipoint, npoint, istep and nstep move to extra.
- many getattr and hasattr calls are replaced by nicer code.
- ArrayTypeCheckDescriptor is replaced by two simple functions.
- Document IOData.charge
- Bug got fixed in poscar format (gvecs)

iodata/formats/cube.py

@@ -187,6 +187,6 @@ def dump_one(f: TextIO, data: IOData):
         attributes.
 
     """
-    title = getattr(data, 'title', 'Created with IOData')
+    title = data.title or 'Created with IOData'
     _write_cube_header(f, title, data.atcoords, data.atnums, data.cube, data.atcorenums)
     _write_cube_data(f, data.cube.data)

iodata/formats/fchk.py

@@ -250,8 +250,8 @@ def load_many(lit: LineIterator) -> Iterator[dict]:
     ------
     out
         Output dictionary containing ``title``, ``atcoords``, ``atnums``,
-        ``atcorenums``, ``ipoint``, ``npoint``, ``istep``, ``nstep``,
-        ``gradient``, ``reaction_coordinate``, and ``energy``.
+        ``atcorenums``, ``extra`` (with ``ipoint``, ``npoint``, ``istep``,
+        ``nstep``), ``gradient``, ``reaction_coordinate``, and ``energy``.
 
     Trajectories from a Gaussian optimization, relaxed scan or IRC calculation
     are written in groups of frames, called "points" in the Gaussian world, e.g.
@@ -294,16 +294,18 @@ def load_many(lit: LineIterator) -> Iterator[dict]:
                 'title': fchk['title'],
                 'atnums': fchk["Atomic numbers"],
                 'atcorenums': fchk["Nuclear charges"],
-                'ipoint': ipoint,
-                'npoint': len(nsteps),
-                'istep': istep,
-                'nstep': nstep,
                 'energy': energy,
                 'atcoords': atcoords,
                 'atgradient': gradients,
+                'extra': {
+                    'ipoint': ipoint,
+                    'npoint': len(nsteps),
+                    'istep': istep,
+                    'nstep': nstep,
+                },
             }
             if prefix == "IRC point":
-                data['reaction_coordinate'] = recor
+                data['extra']['reaction_coordinate'] = recor
             yield data
 
 

iodata/formats/molden.py

@@ -595,7 +595,7 @@ def dump_one(f: TextIO, data: IOData):
     """
     # Print the header
     f.write('[Molden Format]\n')
-    if hasattr(data, 'title'):
+    if data.title is not None:
         f.write('[Title]\n')
         f.write(' {}\n'.format(data.title))
 
@@ -611,8 +611,8 @@ def dump_one(f: TextIO, data: IOData):
     f.write('\n')
 
     # Print the basis set
-    if not hasattr(data, 'obasis'):
-        raise IOError('A Gaussian orbital basis is required to write a molden input file.')
+    if data.obasis is None:
+        raise IOError('A Gaussian orbital basis is required to write a molden file.')
     obasis = data.obasis
 
     # Figure out the pure/Cartesian situation. Note that the Molden

iodata/formats/molpro.py

@@ -135,19 +135,18 @@ def dump_one(f: TextIO, data: IOData):
 
     """
     one_mo = data.one_ints['core_mo']
-    two_mo = data.two_ints['two_mo']
-    nactive = one_mo.shape[0]
-    core_energy = getattr(data, 'core_energy', 0.0)
-    nelec = getattr(data, 'nelec', 0)
-    spinpol = getattr(data, 'spinpol', 0)
 
     # Write header
+    nactive = one_mo.shape[0]
+    nelec = data.nelec or 0
+    spinpol = data.spinpol or 0
     print(f' &FCI NORB={nactive:d},NELEC={nelec:d},MS2={spinpol:d},', file=f)
     print(f"  ORBSYM= {','.join('1' for v in range(nactive))},", file=f)
     print('  ISYM=1', file=f)
     print(' &END', file=f)
 
     # Write integrals and core energy
+    two_mo = data.two_ints['two_mo']
     for i in range(nactive):  # pylint: disable=too-many-nested-blocks
         for j in range(i + 1):
             for k in range(nactive):
@@ -161,5 +160,5 @@ def dump_one(f: TextIO, data: IOData):
             value = one_mo[i, j]
             if value != 0.0:
                 print(f'{value:23.16e} {i+1:4d} {j+1:4d} {0:4d} {0:4d}', file=f)
-    if core_energy != 0.0:
-        print(f'{core_energy:23.16e} {0:4d} {0:4d} {0:4d} {0:4d}', file=f)
+    if data.core_energy is not None:
+        print(f'{data.core_energy:23.16e} {0:4d} {0:4d} {0:4d} {0:4d}', file=f)

iodata/formats/poscar.py

@@ -72,7 +72,7 @@ def dump_one(f: TextIO, data: IOData):
         ``cellvecs`` attributes. It may contain ``title`` attribute.
 
     """
-    print(getattr(data, 'title', 'Created with HORTON'), file=f)
+    print(data.title or 'Created with IOData', file=f)
     print('   1.00000000000000', file=f)
 
     # Write cell vectors, each row is one vector in angstrom:
@@ -90,8 +90,9 @@ def dump_one(f: TextIO, data: IOData):
     print('Direct', file=f)
 
     # Write the coordinates
+    gvecs = np.linalg.inv(data.cellvecs).T
     for uatnum in uatnums:
         indexes = (data.atnums == uatnum).nonzero()[0]
         for index in indexes:
-            row = np.dot(data.gvecs, data.atcoords[index])
+            row = np.dot(gvecs, data.atcoords[index])
             print(f'  {row[0]: 21.16f} {row[1]: 21.16f} {row[2]: 21.16f}   F   F   F', file=f)

iodata/formats/xyz.py

@@ -109,7 +109,7 @@ def dump_one(f: TextIO, data: IOData):
 
     """
     print(data.natom, file=f)
-    print(getattr(data, 'title', 'Created with IODATA module'), file=f)
+    print(data.title or 'Created with IOData', file=f)
     for i in range(data.natom):
         n = num2sym[data.atnums[i]]
         x, y, z = data.atcoords[i] / angstrom