Subdivide vertex data with geometry when converting Marc input to VTK

This commit is contained in:
Brendan Vande Kieft 2018-01-15 17:06:55 -05:00
parent f0511e1ecb
commit ceb0ff09e6
1 changed files with 181 additions and 149 deletions

View File

@ -1,7 +1,7 @@
#!/usr/bin/env python2.7
# -*- coding: UTF-8 no BOM -*-
import os,re
import os,sys,re
import argparse
import damask
import vtk, numpy as np
@ -9,159 +9,191 @@ import vtk, numpy as np
scriptName = os.path.splitext(os.path.basename(__file__))[0]
scriptID = ' '.join([scriptName, damask.version])
parser = argparse.ArgumentParser(description='Convert from Marc input file format to VTK', version = scriptID)
parser.add_argument('filename', type=str, nargs='+', help='files to convert')
parser = argparse.ArgumentParser(description='Convert from Marc input file format (.dat) to VTK format (.vtu)', version = scriptID)
parser.add_argument('filename', type=str, help='file to convert')
parser.add_argument('-t', '--table', type=str, help='ASCIItable file containing nodal data to subdivide and interpolate')
args = parser.parse_args()
files = args.filename
if type(files) is str:
files = [files]
with open(args.filename, 'r') as marcfile:
marctext = marcfile.read();
# Load table (if any)
if args.table is not None:
try:
table = damask.ASCIItable(
name=args.table,
outname='subdivided_{}'.format(args.table),
buffered=True
)
table.head_read()
table.data_readArray()
# Python list is faster for appending
nodal_data = list(table.data)
except: args.table = None
# Extract connectivity chunk from file...
connectivity_text = re.findall(r'connectivity[\n\r]+(.*?)[\n\r]+[a-zA-Z]', marctext, flags=(re.MULTILINE | re.DOTALL))[0]
connectivity_lines = re.split(r'[\n\r]+', connectivity_text, flags=(re.MULTILINE | re.DOTALL))
connectivity_header = connectivity_lines[0]
connectivity_lines = connectivity_lines[1:]
# Construct element map
elements = dict(map(lambda line:
(
int(line[0:10]), # index
{
'type': int(line[10:20]),
'verts': list(map(int, re.split(r' +', line[20:].strip())))
}
), connectivity_lines))
# Extract coordinate chunk from file
coordinates_text = re.findall(r'coordinates[\n\r]+(.*?)[\n\r]+[a-zA-Z]', marctext, flags=(re.MULTILINE | re.DOTALL))[0]
coordinates_lines = re.split(r'[\n\r]+', coordinates_text, flags=(re.MULTILINE | re.DOTALL))
coordinates_header = coordinates_lines[0]
coordinates_lines = coordinates_lines[1:]
# marc input file does not use "e" in scientific notation, this adds it and converts
fl_format = lambda string: float(re.sub(r'(\d)([\+\-])', r'\1e\2', string))
# Construct coordinate map
coordinates = dict(map(lambda line:
(
int(line[0:10]),
np.array([
fl_format(line[10:30]),
fl_format(line[30:50]),
fl_format(line[50:70])
])
), coordinates_lines))
# Subdivide volumes
grid = vtk.vtkUnstructuredGrid()
vertex_count = len(coordinates)
edge_to_vert = dict() # when edges are subdivided, a new vertex in the middle is produced and placed in here
ordered_pair = lambda a, b: (a, b) if a < b else (b, a) # edges are bidirectional
def subdivide_edge(vert1, vert2):
edge = ordered_pair(vert1, vert2)
if edge in edge_to_vert:
return edge_to_vert[edge]
# Vertex does not exist, create it
newvert = len(coordinates) + 1
coordinates[newvert] = 0.5 * (coordinates[vert1] + coordinates[vert2]) # Average
edge_to_vert[edge] = newvert;
# Interpolate nodal data
if args.table is not None:
nodal_data.append(0.5 * (nodal_data[vert1 - 1] + nodal_data[vert2 - 1]))
return newvert;
for el_id in range(1, len(elements) + 1): # Marc starts counting at 1
el = elements[el_id]
if el['type'] == 7:
# Hexahedron, subdivided
# There may be a better way to iterate over these, but this is consistent
# with the ordering scheme provided at https://damask.mpie.de/pub/Documentation/ElementType
subverts = np.zeros((3,3,3), dtype=int)
# Get corners
subverts[0, 0, 0] = el['verts'][0]
subverts[2, 0, 0] = el['verts'][1]
subverts[2, 2, 0] = el['verts'][2]
subverts[0, 2, 0] = el['verts'][3]
subverts[0, 0, 2] = el['verts'][4]
subverts[2, 0, 2] = el['verts'][5]
subverts[2, 2, 2] = el['verts'][6]
subverts[0, 2, 2] = el['verts'][7]
# lower edges
subverts[1, 0, 0] = subdivide_edge(subverts[0, 0, 0], subverts[2, 0, 0])
subverts[2, 1, 0] = subdivide_edge(subverts[2, 0, 0], subverts[2, 2, 0])
subverts[1, 2, 0] = subdivide_edge(subverts[2, 2, 0], subverts[0, 2, 0])
subverts[0, 1, 0] = subdivide_edge(subverts[0, 2, 0], subverts[0, 0, 0])
# middle edges
subverts[0, 0, 1] = subdivide_edge(subverts[0, 0, 0], subverts[0, 0, 2])
subverts[2, 0, 1] = subdivide_edge(subverts[2, 0, 0], subverts[2, 0, 2])
subverts[2, 2, 1] = subdivide_edge(subverts[2, 2, 0], subverts[2, 2, 2])
subverts[0, 2, 1] = subdivide_edge(subverts[0, 2, 0], subverts[0, 2, 2])
# top edges
subverts[1, 0, 2] = subdivide_edge(subverts[0, 0, 2], subverts[2, 0, 2])
subverts[2, 1, 2] = subdivide_edge(subverts[2, 0, 2], subverts[2, 2, 2])
subverts[1, 2, 2] = subdivide_edge(subverts[2, 2, 2], subverts[0, 2, 2])
subverts[0, 1, 2] = subdivide_edge(subverts[0, 2, 2], subverts[0, 0, 2])
# then faces... The edge_to_vert addition is due to there being two ways
# to calculate a face vertex, depending on which opposite vertices are used to subdivide.
# This way, we avoid creating duplicate vertices.
subverts[1, 1, 0] = subdivide_edge(subverts[1, 0, 0], subverts[1, 2, 0])
edge_to_vert[ordered_pair(subverts[0, 1, 0], subverts[2, 1, 0])] = subverts[1, 1, 0]
subverts[1, 0, 1] = subdivide_edge(subverts[1, 0, 0], subverts[1, 0, 2])
edge_to_vert[ordered_pair(subverts[0, 0, 1], subverts[2, 0, 1])] = subverts[1, 0, 1]
subverts[2, 1, 1] = subdivide_edge(subverts[2, 1, 0], subverts[2, 1, 2])
edge_to_vert[ordered_pair(subverts[2, 0, 1], subverts[2, 2, 1])] = subverts[2, 1, 1]
subverts[1, 2, 1] = subdivide_edge(subverts[1, 2, 0], subverts[1, 2, 2])
edge_to_vert[ordered_pair(subverts[0, 2, 1], subverts[2, 2, 1])] = subverts[1, 2, 1]
subverts[0, 1, 1] = subdivide_edge(subverts[0, 1, 0], subverts[0, 1, 2])
edge_to_vert[ordered_pair(subverts[0, 0, 1], subverts[0, 2, 1])] = subverts[0, 1, 1]
subverts[1, 1, 2] = subdivide_edge(subverts[1, 0, 2], subverts[1, 2, 2])
edge_to_vert[ordered_pair(subverts[0, 1, 2], subverts[2, 1, 2])] = subverts[1, 1, 2]
# and finally the center. There are three ways to calculate, but elements should
# not intersect, so the edge_to_vert part isn't needed here.
subverts[1, 1, 1] = subdivide_edge(subverts[1, 1, 0], subverts[1, 1, 2])
for f in files:
with open(f, 'r') as marcfile:
marctext = marcfile.read();
# Extract connectivity chunk from file...
connectivity_text = re.findall(r'connectivity[\n\r]+(.*?)[\n\r]+[a-zA-Z]', marctext, flags=(re.MULTILINE | re.DOTALL))[0]
connectivity_lines = re.split(r'[\n\r]+', connectivity_text, flags=(re.MULTILINE | re.DOTALL))
connectivity_header = connectivity_lines[0]
connectivity_lines = connectivity_lines[1:]
# Construct element map
elements = dict(map(lambda line:
(
int(line[0:10]), # index
{
'type': int(line[10:20]),
'verts': list(map(int, re.split(r' +', line[20:].strip())))
}
), connectivity_lines))
# Extract coordinate chunk from file
coordinates_text = re.findall(r'coordinates[\n\r]+(.*?)[\n\r]+[a-zA-Z]', marctext, flags=(re.MULTILINE | re.DOTALL))[0]
coordinates_lines = re.split(r'[\n\r]+', coordinates_text, flags=(re.MULTILINE | re.DOTALL))
coordinates_header = coordinates_lines[0]
coordinates_lines = coordinates_lines[1:]
# marc input file does not use "e" in scientific notation, this adds it and converts
fl_format = lambda string: float(re.sub(r'(\d)([\+\-])', r'\1e\2', string))
# Construct coordinate map
coordinates = dict(map(lambda line:
(
int(line[0:10]),
np.array([
fl_format(line[10:30]),
fl_format(line[30:50]),
fl_format(line[50:70])
])
), coordinates_lines))
# Now make the hexahedron subelements
# order in which vtk expects vertices for a hexahedron
order = np.array([(0,0,0),(1,0,0),(1,1,0),(0,1,0),(0,0,1),(1,0,1),(1,1,1),(0,1,1)])
for z in range(2):
for y in range(2):
for x in range(2):
hex_ = vtk.vtkHexahedron()
for vert_id in range(8):
coord = order[vert_id] + (x, y, z)
# minus one, since vtk starts at zero but marc starts at one
hex_.GetPointIds().SetId(vert_id, subverts[coord[0], coord[1], coord[2]] - 1)
grid.InsertNextCell(hex_.GetCellType(), hex_.GetPointIds())
# Subdivide volumes
grid = vtk.vtkUnstructuredGrid()
vertex_count = len(coordinates)
edge_to_vert = dict() # when edges are subdivided, a new vertex in the middle is produced and placed in here
ordered_pair = lambda a, b: (a, b) if a < b else (b, a) # edges are bidirectional
else:
damask.util.croak('Unsupported Marc element type: {} (skipping)'.format(el['type']))
def subdivide_edge(vert1, vert2):
edge = ordered_pair(vert1, vert2)
# Load all points
points = vtk.vtkPoints()
for point in range(1, len(coordinates) + 1): # marc indices start at 1
points.InsertNextPoint(coordinates[point].tolist())
if edge in edge_to_vert:
return edge_to_vert[edge]
grid.SetPoints(points)
newvert = len(coordinates) + 1
coordinates[newvert] = 0.5 * (coordinates[vert1] + coordinates[vert2]) # Average
edge_to_vert[edge] = newvert;
return newvert;
# grid now contains the elements from the given marc file
writer = vtk.vtkXMLUnstructuredGridWriter()
writer.SetFileName(re.sub(r'\..+', ".vtu", args.filename)) # *.vtk extension does not work in paraview
if vtk.VTK_MAJOR_VERSION <= 5: writer.SetInput(grid)
else: writer.SetInputData(grid)
writer.Write()
if args.table is not None:
table.info_append([
scriptID + ' ' + ' '.join(sys.argv[1:]),
])
table.head_write()
table.output_flush()
for el_id in range(1, len(elements) + 1):
el = elements[el_id]
if el['type'] == 7:
# Hexahedron, subdivided
table.data = np.array(nodal_data)
# There may be a better way to iterate over these, but this is consistent
# with the ordering scheme provided at https://damask.mpie.de/pub/Documentation/ElementType
table.data_writeArray()
subverts = np.zeros((3,3,3), dtype=int)
# Get corners
subverts[0, 0, 0] = el['verts'][0]
subverts[2, 0, 0] = el['verts'][1]
subverts[2, 2, 0] = el['verts'][2]
subverts[0, 2, 0] = el['verts'][3]
subverts[0, 0, 2] = el['verts'][4]
subverts[2, 0, 2] = el['verts'][5]
subverts[2, 2, 2] = el['verts'][6]
subverts[0, 2, 2] = el['verts'][7]
# lower edges
subverts[1, 0, 0] = subdivide_edge(subverts[0, 0, 0], subverts[2, 0, 0])
subverts[2, 1, 0] = subdivide_edge(subverts[2, 0, 0], subverts[2, 2, 0])
subverts[1, 2, 0] = subdivide_edge(subverts[2, 2, 0], subverts[0, 2, 0])
subverts[0, 1, 0] = subdivide_edge(subverts[0, 2, 0], subverts[0, 0, 0])
# middle edges
subverts[0, 0, 1] = subdivide_edge(subverts[0, 0, 0], subverts[0, 0, 2])
subverts[2, 0, 1] = subdivide_edge(subverts[2, 0, 0], subverts[2, 0, 2])
subverts[2, 2, 1] = subdivide_edge(subverts[2, 2, 0], subverts[2, 2, 2])
subverts[0, 2, 1] = subdivide_edge(subverts[0, 2, 0], subverts[0, 2, 2])
# top edges
subverts[1, 0, 2] = subdivide_edge(subverts[0, 0, 2], subverts[2, 0, 2])
subverts[2, 1, 2] = subdivide_edge(subverts[2, 0, 2], subverts[2, 2, 2])
subverts[1, 2, 2] = subdivide_edge(subverts[2, 2, 2], subverts[0, 2, 2])
subverts[0, 1, 2] = subdivide_edge(subverts[0, 2, 2], subverts[0, 0, 2])
# then faces... The edge_to_vert addition is due to there being two ways
# to calculate a face, depending which opposite vertices are used to subdivide
subverts[1, 1, 0] = subdivide_edge(subverts[1, 0, 0], subverts[1, 2, 0])
edge_to_vert[ordered_pair(subverts[0, 1, 0], subverts[2, 1, 0])] = subverts[1, 1, 0]
subverts[1, 0, 1] = subdivide_edge(subverts[1, 0, 0], subverts[1, 0, 2])
edge_to_vert[ordered_pair(subverts[0, 0, 1], subverts[2, 0, 1])] = subverts[1, 0, 1]
subverts[2, 1, 1] = subdivide_edge(subverts[2, 1, 0], subverts[2, 1, 2])
edge_to_vert[ordered_pair(subverts[2, 0, 1], subverts[2, 2, 1])] = subverts[2, 1, 1]
subverts[1, 2, 1] = subdivide_edge(subverts[1, 2, 0], subverts[1, 2, 2])
edge_to_vert[ordered_pair(subverts[0, 2, 1], subverts[2, 2, 1])] = subverts[1, 2, 1]
subverts[0, 1, 1] = subdivide_edge(subverts[0, 1, 0], subverts[0, 1, 2])
edge_to_vert[ordered_pair(subverts[0, 0, 1], subverts[0, 2, 1])] = subverts[0, 1, 1]
subverts[1, 1, 2] = subdivide_edge(subverts[1, 0, 2], subverts[1, 2, 2])
edge_to_vert[ordered_pair(subverts[0, 1, 2], subverts[2, 1, 2])] = subverts[1, 1, 2]
# and finally the center. There are three ways to calculate, but elements should
# not intersect, so the edge_to_vert part isn't needed here.
subverts[1, 1, 1] = subdivide_edge(subverts[1, 1, 0], subverts[1, 1, 2])
# Now make the hexahedron subelements
# order in which vtk expects vertices for a hexahedron
order = np.array([(0,0,0),(1,0,0),(1,1,0),(0,1,0),(0,0,1),(1,0,1),(1,1,1),(0,1,1)])
for z in range(2):
for y in range(2):
for x in range(2):
hex_ = vtk.vtkHexahedron()
for vert_id in range(8):
coord = order[vert_id] + (x, y, z)
hex_.GetPointIds().SetId(vert_id, subverts[coord[0], coord[1], coord[2]] - 1) # minus one, since vtk starts at zero but marc starts at one
grid.InsertNextCell(hex_.GetCellType(), hex_.GetPointIds())
else:
damask.util.croak('Unsupported Marc element type: {} (skipping)'.format(el['type']))
# Load all points
points = vtk.vtkPoints()
for point in range(1, len(coordinates) + 1): # marc indices start at 1
points.InsertNextPoint(coordinates[point].tolist())
grid.SetPoints(points)
# grid now contains the elements from the given marc file
writer = vtk.vtkXMLUnstructuredGridWriter()
writer.SetFileName(re.sub(r'\..+', ".vtu", f)) # *.vtk extension does not work in paraview
#writer.SetCompressorTypeToZLib()
if vtk.VTK_MAJOR_VERSION <= 5: writer.SetInput(grid)
else: writer.SetInputData(grid)
writer.Write()
table.close()