zoomy-core 0.1.14__py3-none-any.whl

misc/interpolation.py

@@ -0,0 +1,140 @@
+import numpy as np
+from sympy import integrate, diff
+from sympy.abc import x
+from sympy import lambdify
+
+import zoomy_core.mesh.mesh_util as mesh_util
+# from zoomy_core.model.models.shallow_moments import Basis
+
+
+def _find_bounding_element(mesh, position):
+    """
+    Strategy: the faces of the elements are outward facing. If I compute compte the intersection of the normal with the point,
+    resulting in  alpha * normal = minimal_distance; then the alpha needs to be negative for all faces
+    see https://en.wikipedia.org/wiki/Hesse_normal_form
+    """
+    mesh_type = mesh.type
+    for i_elem, vertices in enumerate(mesh.element_vertices):
+        faces = mesh_util._face_order(vertices, mesh_type)
+        face_centers = [
+            mesh_util.center(mesh.vertex_coordinates, np.array(face)) for face in faces
+        ]
+        vector_origin_to_plane = [
+            face_center - position for face_center in face_centers
+        ]
+        face_normals = mesh.element_face_normals[i_elem]
+
+        if _is_point_inside_bounding_faces(face_normals, vector_origin_to_plane):
+            return i_elem
+
+    # outside of domain
+    assert False
+
+
+def _is_point_inside_bounding_faces(outward_face_normals, vectors_OP):
+    for n, p in zip(outward_face_normals, vectors_OP):
+        if np.dot(n, p) < 0:
+            return False
+    return True
+
+
+def to_new_mesh(fields, mesh_old, mesh_new, interp="const", map_fields=None):
+    assert interp == "const"
+
+    fields_new = np.zeros_like(fields)
+
+    for i_elem in range(mesh_new.n_elements):
+        element_center = mesh_new.element_center[i_elem]
+        i_elem_old = _find_bounding_element(mesh_old, element_center)
+        fields_new[i_elem] = fields[i_elem_old]
+    return fields_new
+
+
+# # comute gradients based on FD using scattered pointwise data
+# def compute_gradient_field_2d(points, fields):
+#     def in_hull(points, probe):
+#         n_points = points.shape[0]
+#         n_dim = points.shape[1]
+#         c = np.zeros(n_points)
+#         A = np.r_[points.T, np.ones((1, n_points))]
+#         b = np.r_[probe, np.ones(1)]
+#         lp = linprog(c, A_eq=A, b_eq=b)
+#         return lp.success
+
+#     assert points.shape[1] == 2
+#     grad = np.zeros((fields.shape[0], 2, fields.shape[1]))
+#     eps_x = (points[:, 0].max() - points[:, 0].min()) / 100.0
+#     eps_y = (points[:, 1].max() - points[:, 1].min()) / 100.00
+
+#     # generate evaluation 'stencil' for central differences
+#     xi_0 = np.array(points)
+#     xi_xp = np.array(points)
+#     xi_xp[:, 0] += eps_x
+#     xi_xm = np.array(points)
+#     xi_xm[:, 0] -= eps_x
+#     xi_yp = np.array(points)
+#     xi_yp[:, 1] += eps_y
+#     xi_ym = np.array(points)
+#     xi_ym[:, 1] -= eps_y
+#     factors_x = 2.0 * np.ones((points.shape[0]))
+#     factors_y = 2.0 * np.ones((points.shape[0]))
+#     # correct boundary points with single sided differences
+#     for i in range(xi_xp.shape[0]):
+#         if not in_hull(points, xi_xp[i]):
+#             xi_xp[i, 0] -= eps_x
+#             factors_x[i] = 1.0
+#         if not in_hull(points, xi_xm[i]):
+#             xi_xm[i, 0] += eps_x
+#             factors_x[i] = 1.0
+#         if not in_hull(points, xi_yp[i]):
+#             xi_yp[i, 1] -= eps_y
+#             factors_y[i] = 1.0
+#         if not in_hull(points, xi_ym[i]):
+#             xi_ym[i, 1] += eps_y
+#             factors_y[i] = 1.0
+
+#     for i_field, values in enumerate(fields):
+#         f = griddata(points, values, xi_0)
+#         f_xp = griddata(points, values, xi_xp)
+#         f_xm = griddata(points, values, xi_xm)
+#         f_yp = griddata(points, values, xi_yp)
+#         f_ym = griddata(points, values, xi_ym)
+
+#         dfdx = (f_xp - f_xm) / (factors_x * eps_x + 10 ** (-10))
+#         dfdy = (f_yp - f_ym) / (factors_y * eps_y + 10 ** (-10))
+
+#         grad[i_field, 0, :] = dfdx
+#         grad[i_field, 1, :] = dfdy
+
+#     assert (np.isnan(grad) == False).all()
+#     assert (np.isfinite(grad) == True).all()
+#     return grad
+
+
+# # ps, vs: values at the boundary points
+# # p0, v0, value at the cell_center
+# def compute_gradient(ps, vs, p0, v0, limiter=lambda r: 1.0):
+#     points = np.zeros((ps.shape[0] + 1, 2))
+#     points[:-1, :] = ps[:, :2]
+#     points[-1, :] = p0[:2]
+#     values = np.zeros((vs.shape[0] + 1, vs.shape[1]))
+#     values[:-1, :] = vs
+#     values[-1, :] = v0
+
+#     f = LinearNDInterpolator(points, values)
+#     eps_x = (points[:, 0].max() - points[:, 0].min()) / 100.0
+#     eps_y = (points[:, 1].max() - points[:, 1].min()) / 100.00
+#     x0 = p0[0]
+#     y0 = p0[1]
+
+#     dfdx = (f(x0 + eps_x, y0) - f(x0 - eps_x, y0)) / (2 * eps_x + 10 ** (-10))
+#     rx = (f(x0, y0) - f(x0 - eps_x, y0)) / (f(x0 + eps_x, y0) - f(x0, y0) + 10 ** (-10))
+#     phix = limiter(rx)
+#     dfdy = (f(x0, y0 + eps_y) - f(x0, y0 - eps_y)) / (2 * eps_y + 10 ** (-10))
+#     ry = (f(x0, y0) - f(x0, y0 - eps_y)) / (f(x0, y0 + eps_y) - f(x0, y0) + 10 ** (-10))
+#     phiy = limiter(ry)
+
+#     grad = np.array([phix * dfdx, phiy * dfdy]).T
+#     assert (np.isnan(grad) == False).all()
+#     assert (np.isfinite(grad) == True).all()
+#     return grad

misc/io.py

@@ -0,0 +1,448 @@
+import os
+import numpy as np
+import json
+import shutil
+
+try:
+    import meshio
+
+    _HAVE_MESHIO = True
+except:
+    _HAVE_MESHIO = False
+
+try:
+    import h5py
+
+    _HAVE_H5PY = True
+except ImportError:
+    _HAVE_H5PY = False
+
+# import zoomy_core.mesh.fvm_mesh as fvm_mesh
+from zoomy_core.mesh.mesh import Mesh
+import zoomy_core.mesh.mesh_util as mesh_util
+from zoomy_core.misc.misc import Zstruct, Settings
+from zoomy_core.misc import misc as misc
+from zoomy_core.misc.logger_config import logger
+
+
+def init_output_directory(path, clean):
+    main_dir = misc.get_main_directory()
+
+    path = os.path.join(main_dir, path)
+    os.makedirs(path, exist_ok=True)
+    if clean:
+        filelist = [f for f in os.listdir(path)]
+        for f in filelist:
+            if os.path.isdir(os.path.join(path, f)):
+                shutil.rmtree(os.path.join(path, f))
+            else:
+                os.remove(os.path.join(path, f))
+
+
+def get_hdf5_type(value):
+    out = type(value)
+    if isinstance(value, str):
+        out = h5py.string_dtype()
+    return out
+
+
+def write_dict_to_hdf5(group, d):
+    for key, value in d.items():
+        if isinstance(value, dict):
+            subgroup = group.create_group(key)
+            write_dict_to_hdf5(subgroup, value)
+        elif isinstance(value, (str, int, float, bool)):
+            group.create_dataset(key, data=value, dtype=get_hdf5_type(value))
+        elif isinstance(value, (list, tuple)):
+            group.create_dataset(key, data=value)
+        elif isinstance(value, type(np.ndarray)):
+            group.create_dataset(key, data=value)
+        elif hasattr(value, "as_dict"):
+            subgroup = group.create_group(key)
+            write_dict_to_hdf5(subgroup, value.as_dict())
+        else:
+            logger.warning(f"Skipping unsupported type for key: {
+                           key} -> {type(value)}")
+
+
+def load_hdf5_to_dict(group):
+    d = {}
+    for key, value in group.items():
+        if isinstance(value, h5py.Group):
+            d[key] = load_hdf5_to_dict(value)
+        elif isinstance(value, h5py.Dataset):
+            if value.dtype == h5py.string_dtype():
+                d[key] = value[()].decode("utf-8")
+            else:
+                d[key] = value[()]
+        else:
+            logger.warning(f"Skipping unsupported type for key: {
+                           key} -> {type(value)}")
+
+    return d
+
+
+def save_settings(settings):
+    main_dir = misc.get_main_directory()
+
+    filepath = os.path.join(main_dir, settings.output.directory)
+    with h5py.File(os.path.join(filepath, "settings.h5"), "w") as f:
+        write_dict_to_hdf5(f, settings.as_dict(recursive=True))
+
+
+def load_settings(filepath):
+    main_dir = misc.get_main_directory()
+
+    filepath = os.path.join(main_dir, filepath)
+    with h5py.File(os.path.join(filepath, "settings.h5"), "r") as f:
+        d = load_hdf5_to_dict(f)
+
+    settings = Settings.from_dict(d)
+    return settings
+
+
+def load_settings2(filepath):
+    main_dir = misc.get_main_directory()
+
+    filepath = os.path.join(main_dir, filepath)
+    with h5py.File(os.path.join(filepath, "settings.h5"), "r") as f:
+        model = f["model"]
+        solver = f["solver"]
+        output = f["output"]
+
+        d_model = {}
+        if "parameters" in model:
+            parameters = {k: v[()] for k, v in model["parameters"].items()}
+            parameters = Zstruct(**parameters)
+        for k in model.keys():
+            if k != "parameters":
+                v = model[k][()]
+                if isinstance(v, (str, int, float, bool)):
+                    d_model[k] = v
+                else:
+                    raise ValueError(
+                        f"Unsupported type for model attribute {k}: {type(v)}"
+                    )
+        d_model["parameters"] = parameters
+        model = Zstruct(**d_model)
+        d_solver = {}
+        for k in solver.keys():
+            v = solver[k][()]
+            if isinstance(v, (str, int, float, bool)):
+                d_solver[k] = v
+            else:
+                raise ValueError(
+                    f"Unsupported type for solver attribute {k}: {type(v)}"
+                )
+        solver = Zstruct(**d_solver)
+
+        d_output = {}
+        for k in output.keys():
+            v = output[k][()]
+            if isinstance(v, (str, int, float, bool)):
+                d_output[k] = v
+            else:
+                raise ValueError(
+                    f"Unsupported type for output attribute {k}: {type(v)}"
+                )
+        output = Zstruct(**d_output)
+
+        settings = Settings(model=model, solver=solver, output=output)
+
+        # parameters = {k: v[()] for k, v in f["parameters"].items()}
+        # name = f["name"][()]
+        # output_dir = f["output_dir"][()]
+        # output_snapshots = f["output_snapshots"][()]
+        # output_write_all = f["output_write_all"][()]
+        # output_clean_dir = f["output_clean_dir"][()]
+        # truncate_last_time_step = f["truncate_last_time_step"][()]
+        callbacks = f["callbacks"][()]
+    return settings
+
+
+def clean_files(filepath, filename=".vtk"):
+    main_dir = misc.get_main_directory()
+
+    abs_filepath = os.path.join(main_dir, filepath)
+    if os.path.exists(abs_filepath):
+        for file in os.listdir(abs_filepath):
+            if file.endswith(filename):
+                os.remove(os.path.join(abs_filepath, file))
+
+
+def _save_fields_to_hdf5(filepath, i_snapshot, time, Q, Qaux=None, overwrite=True):
+    i_snap = int(i_snapshot)
+    main_dir = misc.get_main_directory()
+
+    filepath = os.path.join(main_dir, filepath)
+    with h5py.File(filepath, "a") as f:
+        if i_snap == 0 and not "fields" in f.keys():
+            fields = f.create_group("fields")
+        else:
+            fields = f["fields"]
+        group_name = "iteration_" + str(i_snap)
+        if group_name in fields:
+            if overwrite:
+                del fields[group_name]
+            else:
+                raise ValueError(
+                    f"Group {group_name} already exists in {filepath}")
+        attrs = fields.create_group(group_name)
+        attrs.create_dataset("time", data=time, dtype=float)
+        attrs.create_dataset("Q", data=Q)
+        if Qaux is not None:
+            attrs.create_dataset("Qaux", data=Qaux)
+    return i_snapshot + 1.0
+
+
+def get_save_fields_simple(_filepath, write_all, overwrite=True):
+    def _save_hdf5(i_snapshot, time, Q, Qaux):
+        i_snap = int(i_snapshot)
+        main_dir = misc.get_main_directory()
+
+        filepath = os.path.join(main_dir, _filepath)
+
+        with h5py.File(filepath, "a") as f:
+            if i_snap == 0 and not "fields" in f.keys():
+                fields = f.create_group("fields")
+            else:
+                fields = f["fields"]
+            group_name = "iteration_" + str(i_snap)
+            if group_name in fields:
+                if overwrite:
+                    del fields[group_name]
+                else:
+                    raise ValueError(
+                        f"Group {group_name} already exists in {filepath}")
+            attrs = fields.create_group(group_name)
+            attrs.create_dataset("time", data=time, dtype=float)
+            attrs.create_dataset("Q", data=Q)
+            if Qaux is not None:
+                attrs.create_dataset("Qaux", data=Qaux)
+        return i_snapshot + 1.0
+
+    return _save_hdf5
+
+
+def _save_hdf5(_filepath, i_snapshot, time, Q, Qaux, overwrite=True):
+    i_snap = int(i_snapshot)
+    main_dir = misc.get_main_directory()
+
+    filepath = os.path.join(main_dir, _filepath)
+
+    with h5py.File(filepath, "a") as f:
+        if i_snap == 0 and not "fields" in f.keys():
+            fields = f.create_group("fields")
+        else:
+            fields = f["fields"]
+        group_name = "iteration_" + str(i_snap)
+        if group_name in fields:
+            if overwrite:
+                del fields[group_name]
+            else:
+                raise ValueError(
+                    f"Group {group_name} already exists in {filepath}")
+        attrs = fields.create_group(group_name)
+        attrs.create_dataset("time", data=time, dtype=float)
+        attrs.create_dataset("Q", data=Q)
+        if Qaux is not None:
+            attrs.create_dataset("Qaux", data=Qaux)
+    return i_snapshot + 1.0
+
+
+def get_save_fields(_filepath, write_all=False, overwrite=True):
+    if _HAVE_H5PY:
+
+        def save(time, next_write_at, i_snapshot, Q, Qaux):
+            if write_all or time >= next_write_at:
+                return _save_hdf5(
+                    _filepath, i_snapshot, time, Q, Qaux, overwrite=overwrite
+                )
+            else:
+                return i_snapshot
+    else:
+
+        def save(time, next_write_at, i_snapshot, Q, Qaux):
+            if write_all or time >= next_write_at:
+                return i_snapshot + 1
+            else:
+                return i_snapshot
+
+    return save
+
+
+def save_fields_test(a):
+    filepath, time, next_write_at, i_snapshot, Q, Qaux, write_all = a
+    if not write_all and time < next_write_at:
+        return i_snapshot
+
+    _save_fields_to_hdf5(filepath, i_snapshot, time, Q, Qaux)
+    return i_snapshot + 1
+
+
+def load_mesh_from_hdf5(filepath):
+    mesh = Mesh.from_hdf5(filepath)
+    return mesh
+
+
+def load_fields_from_hdf5(filepath, i_snapshot=-1):
+    main_dir = misc.get_main_directory()
+
+    filepath = os.path.join(main_dir, filepath)
+    with h5py.File(filepath, "r") as f:
+        fields = f["fields"]
+        if i_snapshot == -1:
+            i_snapshot = len(fields.keys()) - 1
+        else:
+            i_snapshot = i_snapshot
+        group = fields[f"iteration_{i_snapshot}"]
+        time = group["time"][()]
+        Q = group["Q"][()]
+        Qaux = group["Qaux"][()]
+    return Q, Qaux, time
+
+
+def load_timeline_of_fields_from_hdf5(filepath):
+    main_dir = misc.get_main_directory()
+
+    filepath = os.path.join(main_dir, filepath)
+    l_time = []
+    l_Q = []
+    l_Qaux = []
+    mesh = Mesh.from_hdf5(filepath)
+    with h5py.File(filepath, "r") as f:
+        fields = f["fields"]
+        n_snapshots = len(fields.keys())
+        for i in range(n_snapshots):
+            group = fields[f"iteration_{i}"]
+            time = group["time"][()]
+            Q = group["Q"][()]
+            Qaux = group["Qaux"][()]
+            l_time.append(time)
+            l_Q.append(Q)
+            l_Qaux.append(Qaux)
+    return mesh.cell_centers[0], np.array(l_Q), np.array(l_Qaux), np.array(l_time)
+
+
+def _write_to_vtk_from_vertices_edges(
+    filepath,
+    mesh_type,
+    vertex_coordinates,
+    cell_vertices,
+    fields=None,
+    field_names=None,
+    point_fields=None,
+    point_field_names=None,
+):
+    if not _HAVE_MESHIO:
+        raise RuntimeError(
+            "_write_to_vtk_from_vertices_edges requires meshio, which is not available."
+        )
+    assert (
+        mesh_type == "triangle"
+        or mesh_type == "quad"
+        or mesh_type == "wface"
+        or mesh_type == "hexahedron"
+        or mesh_type == "line"
+        or mesh_type == "tetra"
+    )
+    d_fields = {}
+    n_inner_elements = cell_vertices.shape[0]
+    if fields is not None:
+        if field_names is None:
+            field_names = [str(i) for i in range(fields.shape[0])]
+        for i_fields, _ in enumerate(fields):
+            d_fields[field_names[i_fields]] = [
+                fields[i_fields, :n_inner_elements]]
+    point_d_fields = {}
+    if point_fields is not None:
+        if point_field_names is None:
+            point_field_names = [str(i) for i in range(point_fields.shape[0])]
+        for i_fields, _ in enumerate(point_fields):
+            point_d_fields[point_field_names[i_fields]
+                           ] = point_fields[i_fields]
+    meshout = meshio.Mesh(
+        vertex_coordinates,
+        [(mesh_util.convert_mesh_type_to_meshio_mesh_type(mesh_type), cell_vertices)],
+        cell_data=d_fields,
+        point_data=point_d_fields,
+    )
+    path, filename = os.path.split(filepath)
+    filename_base, filename_ext = os.path.splitext(filename)
+    os.makedirs(path, exist_ok=True)
+    meshout.write(filepath + ".vtk")
+
+
+def generate_vtk(
+    filepath: str,
+    field_names=None,
+    aux_field_names=None,
+    skip_aux=False,
+    filename="out",
+    warp=False,
+):
+    main_dir =  misc.get_main_directory()
+    abs_filepath = os.path.join(main_dir, filepath)
+    path = os.path.dirname(abs_filepath)
+    full_filepath_out = os.path.join(path, filename)
+    # abs_filepath = os.path.join(main_dir, filepath)
+    # with h5py.File(os.path.join(filepath, 'mesh'), "r") as file_mesh, h5py.File(os.path.join(filepath, 'fields'), "r") as file_fields:
+    file = h5py.File(os.path.join(main_dir, filepath), "r")
+    file_fields = file["fields"]
+    mesh = Mesh.from_hdf5(abs_filepath)
+    snapshots = list(file_fields.keys())
+    # init timestamp file
+    vtk_timestamp_file = {"file-series-version": "1.0", "files": []}
+
+    def get_iteration_from_datasetname(name):
+        return int(name.split("_")[1])
+
+    # write out vtk files for each timestamp
+    for snapshot in snapshots:
+        time = file_fields[snapshot]["time"][()]
+        Q = file_fields[snapshot]["Q"][()]
+
+        if not skip_aux:
+            Qaux = file_fields[snapshot]["Qaux"][()]
+        else:
+            Qaux = np.empty((Q.shape[0], 0))
+        output_vtk = f"{filename}.{get_iteration_from_datasetname(snapshot)}"
+
+        # TODO callout to compute pointwise data?
+        point_fields = None
+        point_field_names = None
+
+        if field_names is None:
+            field_names = [str(i) for i in range(Q.shape[0])]
+        if aux_field_names is None:
+            aux_field_names = ["aux_{}".format(
+                str(i)) for i in range(Qaux.shape[0])]
+
+        fields = np.concatenate((Q, Qaux), axis=0)
+        field_names = field_names + aux_field_names
+
+        vertex_coordinates_3d = np.zeros((mesh.vertex_coordinates.shape[1], 3))
+        vertex_coordinates_3d[:, : mesh.dimension] = mesh.vertex_coordinates.T
+
+        _write_to_vtk_from_vertices_edges(
+            os.path.join(path, output_vtk),
+            mesh.type,
+            vertex_coordinates_3d,
+            mesh.cell_vertices.T,
+            fields=fields,
+            field_names=field_names,
+            point_fields=point_fields,
+            point_field_names=point_field_names,
+        )
+
+        vtk_timestamp_file["files"].append(
+            {
+                "name": output_vtk + ".vtk",
+                "time": time,
+            }
+        )
+    # finalize vtk
+    with open(os.path.join(path, f"{filename}.vtk.series"), "w") as f:
+        json.dump(vtk_timestamp_file, f)
+
+    file.close()

misc/logger_config.py

@@ -0,0 +1,18 @@
+
+import os
+import sys
+from loguru import logger
+
+# Remove any default handlers
+logger.remove()
+# Check the ZoomyLog setting
+zoomy_log_mode = os.getenv("ZoomyLog", "Default")
+
+zoomy_log_level = os.getenv("ZoomyLogLevel", "INFO")
+
+main_dir = os.getenv("ZOOMY_DIR", os.getcwd())
+
+if zoomy_log_mode == "Default":
+    logger.add(sys.stderr, level=zoomy_log_level) 
+else:
+    logger.add(os.path.join(main_dir, "logs/log.log"), rotation="1 MB", retention="10 days", compression="zip")