can be done with TSL OIM in current versions

This commit is contained in:
Martin Diehl 2015-05-10 15:56:59 +00:00
parent b60f868a76
commit 6f112b46b5
1 changed files with 0 additions and 99 deletions

View File

@ -1,99 +0,0 @@
#!/usr/bin/env python
# -*- coding: UTF-8 no BOM -*-
import string,os,sys
from optparse import OptionParser
import damask
scriptID = string.replace('$Id$','\n','\\n')
scriptName = os.path.splitext(scriptID.split()[1])[0]
#--------------------------------------------------------------------------------------------------
# MAIN
#--------------------------------------------------------------------------------------------------
parser = OptionParser(option_class=damask.extendableOption, usage='%prog options [file[s]]', description = """
Converts ang files (EBSD Data) from hexagonal grid to a pixel grid
""", version = scriptID)
parser.add_option('-x', dest='columnX', type='int', metavar='int', \
help='column containing x coordinates [%default]')
parser.set_defaults(columnX = 3)
(options,filenames) = parser.parse_args()
counterX = 0
counterY = 0
addPoints = -1 # No of doubled points (must be the same for each odd/even line, initializing with -1 make countin easy!)
#--- setup file handles ---------------------------------------------------------------------------
files = []
if filenames == []:
files.append({'name':'STDIN',
'input':sys.stdin,
'output':sys.stdout,
'croak':sys.stderr,
})
else:
for name in filenames:
if os.path.exists(name):
files.append({'name':name,
'input':open(name),
'output':open(os.path.splitext(name)[0]+'_cub'+os.path.splitext(name)[1], 'w'),
'croak':sys.stdout,
})
#--- loop over input files ------------------------------------------------------------------------
for file in files:
file['croak'].write('\033[1m' + scriptName + '\033[0m: ' + (file['name'] if file['name'] != 'STDIN' else '') + '\n')
x = 0
for line in file['input']:
lineSplit=line.split()
if lineSplit[0]=='#':
if len(lineSplit)>2: # possibly interesting information
if line.split()[2]=='SqrGrid':
file['croak'].write('The file is already a square grid file.')
sys.exit()
if lineSplit[1]=='XSTEP:': stepSizeX = float(lineSplit[2])
if lineSplit[1]=='YSTEP:': stepSizeY = float(lineSplit[2])
if lineSplit[2]=='HexGrid': line='# GRID: SqrGrid\n' # comments are not read by OIM, but still better to be correct
if lineSplit[1]=='NCOLS_ODD:':
NCols = int(int(lineSplit[2])*stepSizeX/stepSizeY)
line='# NCOLS_ODD: %d\n'% NCols
if lineSplit[1]=='NCOLS_EVEN:':
line='# NCOLS_EVEN: %d\n'% NCols
file['output'].write(line)
else: # finished reading of header
xOld = x
x = float(lineSplit[options.columnX]) # current (original) x positions
if x > xOld: # same line, increase X
counterX+=1
else: # new line, increase in Y, reset X
counterY+=1
addPoints = -1 # to start at zero
counterX=0
lineFirstPart ='' # split line around x and y coordinate
for i in xrange(options.columnX):
lineFirstPart =lineFirstPart+' '+lineSplit[i]
lineLastPart =''
for i in xrange(len(lineSplit)- (options.columnX+2)):
lineLastPart =lineLastPart+' '+lineSplit[i+options.columnX+2]
if counterX+addPoints < NCols:
file['output'].write('%s %.6f %.6f %s\n' %(lineFirstPart,(counterX+addPoints)*stepSizeY, # write with new x and y position
counterY*stepSizeY,lineLastPart))
if x - (counterX+addPoints)*stepSizeY > 0.5*stepSizeY and counterX+addPoints+1 < NCols: # double point (interpolation error)
addPoints+=1
file['output'].write('%s %.6f %.6f %s\n' %(lineFirstPart,(counterX+addPoints)*stepSizeY,\
counterY*stepSizeY,lineLastPart))