DAMASK_EICMD/python/tests/test_Result.py

452 lines
20 KiB
Python

import bz2
import pickle
import time
import shutil
import os
import sys
from datetime import datetime
import pytest
import numpy as np
import h5py
from damask import Result
from damask import Rotation
from damask import Orientation
from damask import tensor
from damask import mechanics
from damask import grid_filters
@pytest.fixture
def default(tmp_path,ref_path):
"""Small Result file in temp location for modification."""
fname = '12grains6x7x8_tensionY.hdf5'
shutil.copy(ref_path/fname,tmp_path)
f = Result(tmp_path/fname)
f.view('times',20.0)
return f
@pytest.fixture
def single_phase(tmp_path,ref_path):
"""Single phase Result file in temp location for modification."""
fname = '6grains6x7x8_single_phase_tensionY.hdf5'
shutil.copy(ref_path/fname,tmp_path)
return Result(tmp_path/fname)
@pytest.fixture
def ref_path(ref_path_base):
"""Directory containing reference results."""
return ref_path_base/'Result'
def dict_equal(d1, d2):
for k in d1:
if (k not in d2):
return False
else:
if type(d1[k]) is dict:
return dict_equal(d1[k],d2[k])
else:
if not np.allclose(d1[k],d2[k]):
return False
return True
class TestResult:
def test_self_report(self,default):
print(default)
def test_view_all(self,default):
default.view('increments',True)
a = default.get_dataset_location('F')
default.view('increments','*')
b = default.get_dataset_location('F')
default.view('increments',default.increments_in_range(0,np.iinfo(int).max))
c = default.get_dataset_location('F')
default.view('times',True)
d = default.get_dataset_location('F')
default.view('times','*')
e = default.get_dataset_location('F')
default.view('times',default.times_in_range(0.0,np.inf))
f = default.get_dataset_location('F')
assert a == b == c == d == e ==f
@pytest.mark.parametrize('what',['increments','times','phases']) # ToDo: discuss homogenizations
def test_view_none(self,default,what):
default.view(what,False)
a = default.get_dataset_location('F')
default.view(what,[])
b = default.get_dataset_location('F')
assert a == b == []
@pytest.mark.parametrize('what',['increments','times','phases']) # ToDo: discuss homogenizations
def test_view_more(self,default,what):
default.view(what,False)
default.view_more(what,'*')
a = default.get_dataset_location('F')
default.view(what,True)
b = default.get_dataset_location('F')
assert a == b
@pytest.mark.parametrize('what',['increments','times','phases']) # ToDo: discuss homogenizations
def test_view_less(self,default,what):
default.view(what,True)
default.view_less(what,'*')
a = default.get_dataset_location('F')
default.view(what,False)
b = default.get_dataset_location('F')
assert a == b == []
def test_view_invalid(self,default):
with pytest.raises(AttributeError):
default.view('invalid',True)
def test_add_absolute(self,default):
default.add_absolute('F_e')
loc = {'F_e': default.get_dataset_location('F_e'),
'|F_e|': default.get_dataset_location('|F_e|')}
in_memory = np.abs(default.read_dataset(loc['F_e'],0))
in_file = default.read_dataset(loc['|F_e|'],0)
assert np.allclose(in_memory,in_file)
@pytest.mark.parametrize('mode',['direct','function'])
def test_add_calculation(self,default,tmp_path,mode):
if mode == 'direct':
default.add_calculation('x','2.0*np.abs(#F#)-1.0','-','my notes')
else:
with open(tmp_path/'f.py','w') as f:
f.write("import numpy as np\ndef my_func(field):\n return 2.0*np.abs(field)-1.0\n")
sys.path.insert(0,str(tmp_path))
import f
default.enable_user_function(f.my_func)
default.add_calculation('x','my_func(#F#)','-','my notes')
loc = {'F': default.get_dataset_location('F'),
'x': default.get_dataset_location('x')}
in_memory = 2.0*np.abs(default.read_dataset(loc['F'],0))-1.0
in_file = default.read_dataset(loc['x'],0)
assert np.allclose(in_memory,in_file)
def test_add_stress_Cauchy(self,default):
default.add_stress_Cauchy('P','F')
loc = {'F': default.get_dataset_location('F'),
'P': default.get_dataset_location('P'),
'sigma':default.get_dataset_location('sigma')}
in_memory = mechanics.stress_Cauchy(default.read_dataset(loc['P'],0),
default.read_dataset(loc['F'],0))
in_file = default.read_dataset(loc['sigma'],0)
assert np.allclose(in_memory,in_file)
def test_add_determinant(self,default):
default.add_determinant('P')
loc = {'P': default.get_dataset_location('P'),
'det(P)':default.get_dataset_location('det(P)')}
in_memory = np.linalg.det(default.read_dataset(loc['P'],0)).reshape(-1,1)
in_file = default.read_dataset(loc['det(P)'],0)
assert np.allclose(in_memory,in_file)
def test_add_deviator(self,default):
default.add_deviator('P')
loc = {'P' :default.get_dataset_location('P'),
's_P':default.get_dataset_location('s_P')}
in_memory = tensor.deviatoric(default.read_dataset(loc['P'],0))
in_file = default.read_dataset(loc['s_P'],0)
assert np.allclose(in_memory,in_file)
@pytest.mark.parametrize('eigenvalue,function',[('max',np.amax),('min',np.amin)])
def test_add_eigenvalue(self,default,eigenvalue,function):
default.add_stress_Cauchy('P','F')
default.add_eigenvalue('sigma',eigenvalue)
loc = {'sigma' :default.get_dataset_location('sigma'),
'lambda':default.get_dataset_location(f'lambda_{eigenvalue}(sigma)')}
in_memory = function(tensor.eigenvalues(default.read_dataset(loc['sigma'],0)),axis=1,keepdims=True)
in_file = default.read_dataset(loc['lambda'],0)
assert np.allclose(in_memory,in_file)
@pytest.mark.parametrize('eigenvalue,idx',[('max',2),('mid',1),('min',0)])
def test_add_eigenvector(self,default,eigenvalue,idx):
default.add_stress_Cauchy('P','F')
default.add_eigenvector('sigma',eigenvalue)
loc = {'sigma' :default.get_dataset_location('sigma'),
'v(sigma)':default.get_dataset_location(f'v_{eigenvalue}(sigma)')}
in_memory = tensor.eigenvectors(default.read_dataset(loc['sigma'],0))[:,idx]
in_file = default.read_dataset(loc['v(sigma)'],0)
assert np.allclose(in_memory,in_file)
@pytest.mark.parametrize('d',[[1,0,0],[0,1,0],[0,0,1]])
def test_add_IPF_color(self,default,d):
default.add_IPF_color(d,'O')
loc = {'O': default.get_dataset_location('O'),
'color': default.get_dataset_location('IPFcolor_[{} {} {}]'.format(*d))}
qu = default.read_dataset(loc['O']).view(np.double).squeeze()
crystal_structure = default._get_attribute(default.get_dataset_location('O')[0],'lattice')
c = Orientation(rotation=qu,lattice=crystal_structure)
in_memory = np.uint8(c.IPF_color(np.array(d))*255)
in_file = default.read_dataset(loc['color'])
assert np.allclose(in_memory,in_file)
def test_add_maximum_shear(self,default):
default.add_stress_Cauchy('P','F')
default.add_maximum_shear('sigma')
loc = {'sigma' :default.get_dataset_location('sigma'),
'max_shear(sigma)':default.get_dataset_location('max_shear(sigma)')}
in_memory = mechanics.maximum_shear(default.read_dataset(loc['sigma'],0)).reshape(-1,1)
in_file = default.read_dataset(loc['max_shear(sigma)'],0)
assert np.allclose(in_memory,in_file)
def test_add_Mises_strain(self,default):
t = ['V','U'][np.random.randint(0,2)]
m = np.random.random()*2.0 - 1.0
default.add_strain('F',t,m)
label = f'epsilon_{t}^{m}(F)'
default.add_equivalent_Mises(label)
loc = {label :default.get_dataset_location(label),
label+'_vM':default.get_dataset_location(label+'_vM')}
in_memory = mechanics.equivalent_strain_Mises(default.read_dataset(loc[label],0)).reshape(-1,1)
in_file = default.read_dataset(loc[label+'_vM'],0)
assert np.allclose(in_memory,in_file)
def test_add_Mises_stress(self,default):
default.add_stress_Cauchy('P','F')
default.add_equivalent_Mises('sigma')
loc = {'sigma' :default.get_dataset_location('sigma'),
'sigma_vM':default.get_dataset_location('sigma_vM')}
in_memory = mechanics.equivalent_stress_Mises(default.read_dataset(loc['sigma'],0)).reshape(-1,1)
in_file = default.read_dataset(loc['sigma_vM'],0)
assert np.allclose(in_memory,in_file)
def test_add_Mises_invalid(self,default):
default.add_stress_Cauchy('P','F')
default.add_calculation('sigma_y','#sigma#',unit='y')
default.add_equivalent_Mises('sigma_y')
assert default.get_dataset_location('sigma_y_vM') == []
def test_add_Mises_stress_strain(self,default):
default.add_stress_Cauchy('P','F')
default.add_calculation('sigma_y','#sigma#',unit='y')
default.add_calculation('sigma_x','#sigma#',unit='x')
default.add_equivalent_Mises('sigma_y',kind='strain')
default.add_equivalent_Mises('sigma_x',kind='stress')
loc = {'y' :default.get_dataset_location('sigma_y_vM'),
'x' :default.get_dataset_location('sigma_x_vM')}
assert not np.allclose(default.read_dataset(loc['y'],0),default.read_dataset(loc['x'],0))
def test_add_norm(self,default):
default.add_norm('F',1)
loc = {'F': default.get_dataset_location('F'),
'|F|_1':default.get_dataset_location('|F|_1')}
in_memory = np.linalg.norm(default.read_dataset(loc['F'],0),ord=1,axis=(1,2),keepdims=True)
in_file = default.read_dataset(loc['|F|_1'],0)
assert np.allclose(in_memory,in_file)
def test_add_stress_second_Piola_Kirchhoff(self,default):
default.add_stress_second_Piola_Kirchhoff('P','F')
loc = {'F':default.get_dataset_location('F'),
'P':default.get_dataset_location('P'),
'S':default.get_dataset_location('S')}
in_memory = mechanics.stress_second_Piola_Kirchhoff(default.read_dataset(loc['P'],0),
default.read_dataset(loc['F'],0))
in_file = default.read_dataset(loc['S'],0)
assert np.allclose(in_memory,in_file)
@pytest.mark.skip(reason='requires rework of lattice.f90')
@pytest.mark.parametrize('polar',[True,False])
def test_add_pole(self,default,polar):
pole = np.array([1.,0.,0.])
default.add_pole('O',pole,polar)
loc = {'O': default.get_dataset_location('O'),
'pole': default.get_dataset_location('p^{}_[1 0 0)'.format(u'' if polar else 'xy'))}
rot = Rotation(default.read_dataset(loc['O']).view(np.double))
rotated_pole = rot * np.broadcast_to(pole,rot.shape+(3,))
xy = rotated_pole[:,0:2]/(1.+abs(pole[2]))
in_memory = xy if not polar else \
np.block([np.sqrt(xy[:,0:1]*xy[:,0:1]+xy[:,1:2]*xy[:,1:2]),np.arctan2(xy[:,1:2],xy[:,0:1])])
in_file = default.read_dataset(loc['pole'])
assert np.allclose(in_memory,in_file)
def test_add_rotation(self,default):
default.add_rotation('F')
loc = {'F': default.get_dataset_location('F'),
'R(F)': default.get_dataset_location('R(F)')}
in_memory = mechanics.rotation(default.read_dataset(loc['F'],0)).as_matrix()
in_file = default.read_dataset(loc['R(F)'],0)
assert np.allclose(in_memory,in_file)
def test_add_spherical(self,default):
default.add_spherical('P')
loc = {'P': default.get_dataset_location('P'),
'p_P': default.get_dataset_location('p_P')}
in_memory = tensor.spherical(default.read_dataset(loc['P'],0),False).reshape(-1,1)
in_file = default.read_dataset(loc['p_P'],0)
assert np.allclose(in_memory,in_file)
def test_add_strain(self,default):
t = ['V','U'][np.random.randint(0,2)]
m = np.random.random()*2.0 - 1.0
default.add_strain('F',t,m)
label = f'epsilon_{t}^{m}(F)'
loc = {'F': default.get_dataset_location('F'),
label: default.get_dataset_location(label)}
in_memory = mechanics.strain(default.read_dataset(loc['F'],0),t,m)
in_file = default.read_dataset(loc[label],0)
assert np.allclose(in_memory,in_file)
def test_add_stretch_right(self,default):
default.add_stretch_tensor('F','U')
loc = {'F': default.get_dataset_location('F'),
'U(F)': default.get_dataset_location('U(F)')}
in_memory = mechanics.stretch_right(default.read_dataset(loc['F'],0))
in_file = default.read_dataset(loc['U(F)'],0)
assert np.allclose(in_memory,in_file)
def test_add_stretch_left(self,default):
default.add_stretch_tensor('F','V')
loc = {'F': default.get_dataset_location('F'),
'V(F)': default.get_dataset_location('V(F)')}
in_memory = mechanics.stretch_left(default.read_dataset(loc['F'],0))
in_file = default.read_dataset(loc['V(F)'],0)
assert np.allclose(in_memory,in_file)
def test_add_invalid(self,default):
with pytest.raises(TypeError):
default.add_calculation('#invalid#*2')
@pytest.mark.parametrize('overwrite',['off','on'])
def test_add_overwrite(self,default,overwrite):
default.view('times',default.times_in_range(0,np.inf)[-1])
default.add_stress_Cauchy()
loc = default.get_dataset_location('sigma')
with h5py.File(default.fname,'r') as f:
# h5py3 compatibility
try:
created_first = f[loc[0]].attrs['created'].decode()
except AttributeError:
created_first = f[loc[0]].attrs['created']
created_first = datetime.strptime(created_first,'%Y-%m-%d %H:%M:%S%z')
if overwrite == 'on':
default.allow_modification()
else:
default.disallow_modification()
time.sleep(2.)
try:
default.add_calculation('sigma','#sigma#*0.0+311.','not the Cauchy stress')
except ValueError:
pass
with h5py.File(default.fname,'r') as f:
# h5py3 compatibility
try:
created_second = f[loc[0]].attrs['created'].decode()
except AttributeError:
created_second = f[loc[0]].attrs['created']
created_second = datetime.strptime(created_second,'%Y-%m-%d %H:%M:%S%z')
if overwrite == 'on':
assert created_first < created_second and np.allclose(default.read_dataset(loc),311.)
else:
assert created_first == created_second and not np.allclose(default.read_dataset(loc),311.)
@pytest.mark.parametrize('allowed',['off','on'])
def test_rename(self,default,allowed):
if allowed == 'on':
F = default.read_dataset(default.get_dataset_location('F'))
default.allow_modification()
default.rename('F','new_name')
assert np.all(F == default.read_dataset(default.get_dataset_location('new_name')))
default.disallow_modification()
with pytest.raises(PermissionError):
default.rename('P','another_new_name')
@pytest.mark.parametrize('mode',['cell','node'])
def test_coordinates(self,default,mode):
if mode == 'cell':
a = grid_filters.coordinates0_point(default.cells,default.size,default.origin)
b = default.coordinates0_point.reshape(tuple(default.cells)+(3,),order='F')
elif mode == 'node':
a = grid_filters.coordinates0_node(default.cells,default.size,default.origin)
b = default.coordinates0_node.reshape(tuple(default.cells+1)+(3,),order='F')
assert np.allclose(a,b)
@pytest.mark.parametrize('output',['F',[],['F','P']])
def test_vtk(self,tmp_path,default,output):
os.chdir(tmp_path)
default.save_VTK(output)
@pytest.mark.parametrize('mode',['point','cell'])
def test_vtk_mode(self,tmp_path,single_phase,mode):
os.chdir(tmp_path)
single_phase.save_VTK(mode=mode)
def test_XDMF(self,tmp_path,single_phase,update,ref_path):
for shape in [('scalar',()),('vector',(3,)),('tensor',(3,3)),('matrix',(12,))]:
for dtype in ['f4','f8','i1','i2','i4','i8','u1','u2','u4','u8']:
single_phase.add_calculation(f'{shape[0]}_{dtype}',f"np.ones(np.shape(#F#)[0:1]+{shape[1]},'{dtype}')")
fname = os.path.splitext(os.path.basename(single_phase.fname))[0]+'.xdmf'
os.chdir(tmp_path)
single_phase.save_XDMF()
if update:
shutil.copy(tmp_path/fname,ref_path/fname)
assert sorted(open(tmp_path/fname).read()) == sorted(open(ref_path/fname).read()) # XML is not ordered
def test_XDMF_invalid(self,default):
with pytest.raises(TypeError):
default.save_XDMF()
@pytest.mark.parametrize('view,output,compress,strip',
[({},['F','P','F','L_p','F_e','F_p'],True,True),
({'increments':3},'F',True,True),
({'increments':[1,8,3,4,5,6,7]},['F','P'],True,True),
({'phases':['A','B']},['F','P'],True,True),
({'phases':['A','C'],'homogenizations':False},['F','P','O'],True,True),
({'phases':False,'homogenizations':False},['F','P','O'],True,True),
({'phases':False},['Delta_V'],True,True),
({},['u_p','u_n'],False,False)],
ids=list(range(8)))
def test_read(self,update,request,ref_path,view,output,compress,strip):
result = Result(ref_path/'4grains2x4x3_compressionY.hdf5')
for key,value in view.items():
result.view(key,value)
N = request.node.name[8:].split('[')[1].split(']')[0]
cur = result.read(output,compress,strip)
if update:
with bz2.BZ2File(ref_path/f'read_{N}.pbz2','w') as f:
pickle.dump(cur,f)
with bz2.BZ2File(ref_path/f'read_{N}.pbz2') as f:
assert dict_equal(cur,pickle.load(f))
@pytest.mark.parametrize('view,output,compress,constituents,strip',
[({},['F','P','F','L_p','F_e','F_p'],True,True,None),
({'increments':3},'F',True,True,[0,1,2,3,4,5,6,7]),
({'increments':[1,8,3,4,5,6,7]},['F','P'],True,True,1),
({'phases':['A','B']},['F','P'],True,True,[1,2]),
({'phases':['A','C'],'homogenizations':False},['F','P','O'],True,True,[0,7]),
({'phases':False,'homogenizations':False},['F','P','O'],True,True,[1,2,3,4]),
({'phases':False},['Delta_V'],True,True,[1,2,4]),
({},['u_p','u_n'],False,False,None)],
ids=list(range(8)))
def test_place(self,update,request,ref_path,view,output,compress,strip,constituents):
result = Result(ref_path/'4grains2x4x3_compressionY.hdf5')
for key,value in view.items():
result.view(key,value)
N = request.node.name[8:].split('[')[1].split(']')[0]
cur = result.place(output,compress,strip,constituents)
if update:
with bz2.BZ2File(ref_path/f'place_{N}.pbz2','w') as f:
pickle.dump(cur,f)
with bz2.BZ2File(ref_path/f'place_{N}.pbz2') as f:
assert dict_equal(cur,pickle.load(f))