class Lagrangian:
"""
This class provides pure Lagrangian kernels.
The particle just follows the local velocity flow field.
Attributes:
-Vi (function): The interpolant function, to evaluate the
local flow velocity field V(r(t),t).
"""
def __init__(self,json_dict):
"""
Lagrangian kernel constructor.
Args:
Vi (function): Interpolant function.
"""
self.Vi = InputFields()
self.Vi.get_info(json_dict['velocity'])
self.Vi.get_mfds()
self.Vi.get_grid()
self.Vi.get_interpolants()
self.boundaries = []
def F(self, r, t):
"""
Model Kernel function to send to the solver module.
Args:
r (array): Array containing the position of the particles [space_units].
t (float): Time instant [time_units]
Returns:
array: Array containing the velocity componentes evaluted. [space_units/time_units]
"""
if self.boundaries:
r = self.boundaries.F(r)
return self.Vi.F(r, t)
class LagrangianSpherical2D:
"""
This class provides Lagrangian Kernels, for spherical (lat,lon) integrations,
considering a perfectly spherical earth with a 6370Km radius.
The particle just follows the local flow velocity field.
Attributes:
- m_to_deg (float): Conversion from [m/s] to [degrees/s].
- Vi (function): The interpolant function, to evaluate the local flow velocity field V(r(t),t).
"""
def __init__(self, json_dict):
"""
Lagrangian kernel constructor.
Args:
Vi (function): Interpolant function.
"""
self.m_to_deg = (np.pi/180.)*6370000.
self.Vi = InputFields()
self.Vi.get_info(json_dict['velocity'])
self.Vi.get_mfds()
self.Vi.get_grid()
self.Vi.get_interpolants()
def F(self, r, t):
"""
Model Kernel function to send to the solver module.
Args:
- r (array): Array containing the position of the particles [space_units].
- t (float): Time instant [time_units].
Returns:
- array: Array containing the velocity componentes evaluted [space_units/time_units].
"""
drdt = self.Vi.F(r, t)
drdt[:,0] = drdt[:,0]/(self.m_to_deg*np.cos((np.pi/180.)*r[:,1]))
drdt[:,1] = drdt[:,1]/self.m_to_deg
return drdt
class InitialConditions:
"""
This module creates regular meshes of multidimensional initial conditions,
and turn it into arrays of points, using the order: [variable, solution_id].
Example: Given 10 Lagrangian particles in cartesian coordinates. we require x,y,z position
to describe its state at t time, so the resulting array will be: [3,10].
Using this,
::
r0[1]
we take the x-component for all the particles.
Also, if a mask is provided, the points can be masked and suppressed in
order to create regular initial conditions with any shape considered.
Future: The idea is to increase its functionality by adding new functions for
multidimensional problems and non-regular geometries.
Attributes:
- r0 (np.array): An array of initial points.
- r0_raw (list): Copy of the original array with a mask included, to
- create a np.masked array.
- mask (bool): Array of booleans to select the valid points to compute.
- span (np.array): The initial grid of points to generate the initial conditions domain.
"""
def __init__(self):
"""
Init the object to create the initial conditions.
Args:
- **r0_dict: The information provided by the dictionary "r0" in the setup JSON file.
"""
self.setup = []
self.names = []
self.ranges = []
self.step = []
self.span = []
self.mask = []
self.maski = []
self.r0 = []
self.r0_raw = []
self.t0 = []
self.time_range = []
self.dt = []
self.r = []
self.t = []
self.t_fmt = []
self.t0_fmt = []
self.dt_fmt = []
self.ds = []
self.coords_names = []
self.tspan = []
self.tspan_fmt = []
self.parameters = []
def read_config(self, json_dict):
self.names = json_dict['names']
if 'file' in json_dict:
self.file = json_dict['file']
else:
self.ranges = json_dict['ranges']
self.step = json_dict['step']
self.time_range = json_dict['time_range']
if 'reference_time' in json_dict:
self.time_ref = json_dict['reference_time']
if 'parameters' in json_dict:
self.parameters = json_dict['parameters']
return
def spanning(self):
"""
Turns the information provided by the by the dictionary "r0" in the setup JSON file,
regarding variables into arrays of initial conditions to define the initial mesh axis.
Returns:
- list: List of np.arrays.
"""
for k in range(0, len(self.names)):
self.span.append(
np.arange(self.ranges[k][0], self.ranges[k][1], self.step[k]))
return
def set_array(self):
"""
It creates the array of initial conditions.
Returns:
- array: np.array of initial conditions in the form [varible, particle_id]
"""
mesh = np.meshgrid(*self.span, indexing='ij')
# here we use indexing ij, this is to work with the correspondence
# x,y,z --> i,j,k indexing. Then, when the netcdf is writen, this is
# tranposed in order to keep the netcdf convection (k,j,i indexing)
self.coords_names = [name + "_0" for name in self.names]
nvars = len(self.names)
coords = dict(zip(self.coords_names, zip(self.coords_names,
self.span)))
variables = dict(
zip(self.names, zip(nvars * [self.coords_names], mesh)))
self.ds = xr.Dataset(variables, coords=coords)
self.ds = self.ds.expand_dims(dim='time', axis=len(self.coords_names))
for var in self.names:
self.ds[var].values.setflags(write=True)
self.r0 = np.stack((list(map(np.ravel, mesh))))
self.r0 = self.r0.transpose()
return
def set_mask(self, json_dict):
self.maski = InputFields()
self.maski.get_info(json_dict['mask'])
self.maski.get_mfds()
self.maski.get_grid()
self.maski.get_interpolants(method='nearest')
def apply_mask(self):
"""
Apply the mask to the points, and filter the non-true points.
It also generates a copy of the array with the mask included (numpy masked array)
to reconstruct the initial array, in the postprocessing stage.
WARNING::
The masked array cannot be passed directly to computations (it is heavily slow).
"""
print('\n')
print('*************************************')
print('LAGAR InitialConditions Information ')
print('*************************************')
print('Info: True condition => Point masked by position')
print('Info: Points True condition => not integrate')
print('Points pre-mask:', self.r0.shape[0])
print('Stage: Appliying mask filter points')
self.mask = self.maski.F_s(self.r0) == 1
# Added tuple to avoid the Future Warning.
remove_mask = (~self.mask).any(axis=1)
self.r0 = self.r0[(remove_mask)]
if self.ds:
self.ds['mask'] = (self.coords_names,
self.mask.reshape(self.get_dims()))
print('Points post-mask:', self.r0.shape[0])
print('\n')
return
def read_points_file(self, delimiter=';'):
"""
Read the information of initial conditions from a CSV file.
Args:
delimiter (str, optional): delimiter for CSV files.
Returns:
void: sets the array with the CSV readed values.
"""
if self.file[-4:] == '.csv':
csv = pd.read_csv(self.file, delimiter=delimiter)
self.r0 = csv[self.names].values
elif self.file[-3:] == '.nc':
ds = xr.open_dataset(self.file)
try:
self.r0 = np.stack([ds[name].values for name in self.names],
axis=1)
except:
self.r0 = np.array([ds[name].values for name in self.names])
return
def span_time(self):
if isinstance(self.time_range[0], str):
self.tspan_fmt = pd.date_range(start=self.time_range[0],
end=self.time_range[1],
freq=self.time_range[2])
self.reference_time = pd.datetime.strptime(self.time_ref,
'%Y-%m-%d %H:%M:%S')
self.dt_fmt = self.time_range[2]
self.t0_fmt = self.tspan_fmt[0]
self.tspan = (self.tspan_fmt - self.reference_time
).astype('timedelta64[s]').astype('f8').values
self.t0 = self.tspan[0]
self.dt = float(self.time_range[2][0:-1])
print('+++++++++++++++++++++++++++++++')
print('LAGAR InitialCondtions time')
print('+++++++++++++++++++++++++++++++')
print('Reference_time:', self.reference_time)
print('Number of steps:', self.tspan)
print('Time step used:', self.dt)
print('Date started:', self.t0_fmt)
print('Date end:', self.tspan_fmt[-1])
print('Date started in seconds', self.t0)
print('Date end in seconds:', self.tspan_fmt[-1])
else:
self.tspan = np.arange(self.time_range[0], self.time_range[1],
self.time_range[2])
self.t0 = self.tspan[0]
self.dt = self.time_range[2]
print('+++++++++++++++++++++++++++++++')
print('LAGAR InitialCondtions time')
print('+++++++++++++++++++++++++++++++')
print('Number of steps:', self.tspan.size)
print('Time step used:', self.dt)
print('Date started in seconds', self.t0)
return
def get_dims(self):
"""
Get the dimensions of the initial conditions domain.
Returns:
list: the list containing the dimensions of the initial conditions domain.
"""
return list(map(np.size, self.span))
def generate(self, json_dict):
self.read_config(json_dict)
if 'file' in json_dict:
self.read_points_file()
else:
self.spanning()
self.set_array()
if 'mask' in json_dict:
self.set_mask(json_dict)
self.apply_mask()
self.span_time()
def update(self, json_dict):
self.read_config(json_dict)
self.span_time()