import os.path as op
import numpy as nm
import tables as pt
from sfepy.base.base import output, Struct
from sfepy.base.conf import ProblemConf, get_standard_keywords
from sfepy.homogenization.homogen_app import HomogenizationApp
from sfepy.homogenization.engine import HomogenizationEngine
from sfepy.homogenization.coefficients import Coefficients
from sfepy.discrete import Problem
from sfepy.discrete.fem.meshio import HDF5MeshIO
homogen_app_cache = {}
[docs]
def get_micro_conf(micro_filename, define_args, output_dir):
required, other = get_standard_keywords()
required.remove('equations')
conf = ProblemConf.from_file(micro_filename, required, other,
verbose=False, define_args=define_args)
if output_dir is not None:
conf.options.output_dir = output_dir
return conf
[docs]
def coefs_to_dict(coefs, mode, n):
out = {}
if mode == None:
for key, val in coefs.__dict__.items():
out[key] = val
elif mode == 'qp':
for key, val in coefs.__dict__.items():
if type( val ) == nm.ndarray or type(val) == nm.float64:
out[key] = nm.tile( val, (n, 1, 1) )
elif type(val) == dict:
for key2, val2 in val.items():
if type(val2) == nm.ndarray or type(val2) == nm.float64:
out[key+'_'+key2] = nm.tile(val2, (n, 1, 1))
else:
out = None
return out
[docs]
def get_homog_coefs_linear(ts, coor, mode,
micro_filename=None, regenerate=False,
coefs_filename=None, define_args=None,
output_dir=None):
nc = None if coor is None else coor.shape[0]
if micro_filename in homogen_app_cache:
return coefs_to_dict(homogen_app_cache[micro_filename][0], mode, nc)
oprefix = output.prefix
output.prefix = 'micro:'
conf = get_micro_conf(micro_filename, define_args, output_dir)
if coefs_filename is None:
coefs_filename = conf.options.get('coefs_filename', 'coefs')
coefs_filename = op.join(conf.options.get('output_dir', '.'),
coefs_filename) + '.h5'
if not regenerate:
if op.exists( coefs_filename ):
if not pt.is_hdf5_file( coefs_filename ):
regenerate = True
else:
regenerate = True
recovery_hook = conf.options.get('recovery_hook', None)
if regenerate:
options = Struct( output_filename_trunk = None )
app = HomogenizationApp( conf, options, 'micro:' )
coefs, deps = app(ret_all=True)
if type(coefs) is tuple:
coefs = coefs[0]
coefs.to_file_hdf5( coefs_filename )
rhook = (None if recovery_hook is None else
conf.get_function(recovery_hook))
homogen_app_cache[micro_filename] = (coefs, app.he, deps, rhook)
else:
coefs = Coefficients.from_file_hdf5( coefs_filename )
recovery_hook = conf.options.get('recovery_hook', None)
if recovery_hook is not None:
corrs = get_correctors_from_file_hdf5(dump_names=coefs.save_names)
rhook = conf.get_function(recovery_hook)
problem = Problem.from_conf(conf, init_equations=False,
init_solvers=False)
options = Struct(output_filename_trunk=None)
he = HomogenizationEngine(problem, options)
homogen_app_cache[micro_filename] = (coefs, he, corrs, rhook)
else:
homogen_app_cache[micro_filename] = (coefs, None, None, None)
output.prefix = oprefix
return coefs_to_dict(coefs, mode, nc)
[docs]
def get_homog_coefs_nonlinear(ts, coor, mode, macro_data=None,
term=None, problem=None,
iteration=None, define_args=None,
output_dir=None, ret_corrs=False, **kwargs):
if not (mode == 'qp'):
return
oprefix = output.prefix
output.prefix = 'micro:'
micro_filename = problem.conf.options.micro_filename
if micro_filename not in homogen_app_cache:
conf = get_micro_conf(micro_filename, define_args, output_dir)
options = Struct(output_filename_trunk=None)
app = HomogenizationApp(conf, options, 'micro:', n_micro=coor.shape[0])
homogen_app_cache[micro_filename] = app
else:
app = homogen_app_cache[micro_filename]
if macro_data is not None:
macro_data['macro_time_step'] = ts.step
app.setup_macro_data(macro_data)
coefs, deps = app(ret_all=True, itime=ts.step, iiter=iteration)
if type(coefs) is tuple:
coefs = coefs[0]
out = {}
for key, val in coefs.__dict__.items():
if isinstance(val, list):
out[key] = nm.array(val)
elif isinstance(val, dict):
for key2, val2 in val.items():
out[key+'_'+key2] = nm.array(val2)
for key in out.keys():
shape = out[key].shape
if len(shape) == 1:
out[key] = out[key].reshape(shape + (1, 1))
elif len(shape) == 2:
out[key] = out[key].reshape(shape + (1,))
output.prefix = oprefix
if ret_corrs:
return out, deps
else:
return out
[docs]
def get_correctors_from_file_hdf5(coefs_filename='coefs.h5',
dump_names=None):
if dump_names == None:
coefs = Coefficients.from_file_hdf5( coefs_filename )
if hasattr(coefs, 'save_names'):
dump_names = coefs.save_names
else:
raise ValueError( ' "filenames" coefficient must be used!' )
out = {}
for key, val in dump_names.items():
if type(val) in [tuple, list]:
h5name, corr_name = val
else:
h5name, corr_name = val, op.split(val)[-1]
io = HDF5MeshIO(h5name + '.h5')
try:
ts = io.read_time_stepper()
except ValueError:
ts = None
if ts is None:
data = io.read_data(0)
dkeys = list(data.keys())
corr = {}
for dk in dkeys:
corr[dk] = data[dk].data.reshape(data[dk].shape)
out[corr_name] = corr
else:
n_step = ts[3]
out[corr_name] = []
for step in range(n_step):
data = io.read_data(step)
dkeys = list(data.keys())
corr = {}
for dk in dkeys:
corr[dk] = data[dk].data.reshape(data[dk].shape)
out[corr_name].append(corr)
out[corr_name + '_ts'] = ts
return out