DAMASK_EICMD/processing/post/scaleData.py

109 lines
5.7 KiB
Python
Executable File

#!/usr/bin/env python
# -*- coding: UTF-8 no BOM -*-
import os,sys,string
import numpy as np
from collections import defaultdict
from optparse import OptionParser
import damask
scriptID = '$Id$'
scriptName = os.path.splitext(scriptID.split()[1])[0]
# --------------------------------------------------------------------
# MAIN
# --------------------------------------------------------------------
parser = OptionParser(option_class=damask.extendableOption, usage='%prog options [file[s]]', description = """
Uniformly scale values in scalar/special, vector, or tensor columns by given factor.
""", version = scriptID)
parser.add_option('-s','--special', dest='special', action='extend', metavar='<string LIST>',
help='heading of columns containing field values of special dimension')
parser.add_option('-d','--dimension',dest='N', type='int', metavar='int',
help='dimension of special field values [%default]')
parser.add_option('-v','--vector', dest='vector', action='extend', metavar='<string LIST>',
help='column heading of vector to scale')
parser.add_option('-t','--tensor', dest='tensor', action='extend', metavar='<string LIST>',
help='column heading of tensor to scale')
parser.add_option('-f','--factor', dest='factor', action='extend', metavar='<float LIST>',
help='list of scalar/special, vector, and tensor scaling factors (in this order!)')
parser.set_defaults(special = [])
parser.set_defaults(vector = [])
parser.set_defaults(tensor = [])
parser.set_defaults(factor = [])
parser.set_defaults(N = 1)
(options,filenames) = parser.parse_args()
options.factor = np.array(options.factor,'d')
datainfo = { # list of requested labels per datatype
'special': {'len':options.N,
'label':[]},
'vector': {'len':3,
'label':[]},
'tensor': {'len':9,
'label':[]},
}
length = 0
if options.special != []: datainfo['special']['label'] += options.special; length += len(options.special)
if options.vector != []: datainfo['vector']['label'] += options.vector; length += len(options.vector)
if options.tensor != []: datainfo['tensor']['label'] += options.tensor; length += len(options.tensor)
if len(options.factor) != length:
parser.error('length of scaling vector does not match column count...')
# --- loop over input files -------------------------------------------------------------------------
if filenames == []:
filenames = ['STDIN']
for name in filenames:
if name == 'STDIN':
file = {'name':'STDIN', 'input':sys.stdin, 'output':sys.stdout, 'croak':sys.stderr}
file['croak'].write('\033[1m'+scriptName+'\033[0m\n')
else:
if not os.path.exists(name): continue
file = {'name':name, 'input':open(name), 'output':open(name+'_tmp','w'), 'croak':sys.stderr}
file['croak'].write('\033[1m'+scriptName+'\033[0m: '+file['name']+'\n')
table = damask.ASCIItable(file['input'],file['output'],buffered=False) # make unbuffered ASCII_table
table.head_read() # read ASCII header info
table.info_append(scriptID + '\t' + ' '.join(sys.argv[1:]))
# --------------- figure out columns to process ---------------------------------------------------
active = defaultdict(list)
column = defaultdict(dict)
for datatype,info in datainfo.items():
for label in info['label']:
key = '1_'+label if info['len'] > 1 else label
if key in table.labels:
active[datatype].append(label)
column[datatype][label] = table.labels.index(key) # remember columns of requested data
else:
file['croak'].write('column %s not found...\n'%label)
# ------------------------------------------ assemble header ---------------------------------------
table.head_write()
# ------------------------------------------ process data ------------------------------------------
outputAlive = True
while outputAlive and table.data_read(): # read next data line of ASCII table
i = 0
for datatype,labels in sorted(active.items(),key=lambda x:datainfo[x[0]]['len']): # loop over special,vector,tensor
for label in labels: # loop over all requested labels
for j in xrange(datainfo[datatype]['len']): # loop over entity elements
table.data[column[datatype][label]+j] = float(table.data[column[datatype][label]+j]) * options.factor[i]
i += 1
outputAlive = table.data_write() # output processed line
# ------------------------------------------ output result -----------------------------------------
outputAlive and table.output_flush() # just in case of buffered ASCII table
table.input_close() # close input ASCII table (works for stdin)
table.output_close() # close output ASCII table (works for stdout)
if file['name'] != 'STDIN':
os.rename(file['name']+'_tmp',file['name']) # overwrite old one with tmp new