unmaintained python2.7 scripts
This commit is contained in:
parent
bb90539f7c
commit
d87d13087c
|
@ -1,146 +0,0 @@
|
|||
#!/usr/bin/env python2.7
|
||||
|
||||
import os
|
||||
import sys
|
||||
from optparse import OptionParser
|
||||
|
||||
import numpy as np
|
||||
|
||||
import damask
|
||||
|
||||
|
||||
scriptName = os.path.splitext(os.path.basename(__file__))[0]
|
||||
scriptID = ' '.join([scriptName,damask.version])
|
||||
|
||||
|
||||
# --------------------------------------------------------------------
|
||||
# MAIN
|
||||
# --------------------------------------------------------------------
|
||||
|
||||
parser = OptionParser(option_class=damask.extendableOption, usage='%prog options [file[s]]', description = """
|
||||
Generate histogram of N bins in given data range.
|
||||
|
||||
""", version = scriptID)
|
||||
|
||||
parser.add_option('-d','--data',
|
||||
dest = 'data',
|
||||
type = 'string', metavar = 'string',
|
||||
help = 'column heading for data')
|
||||
parser.add_option('-w','--weights',
|
||||
dest = 'weights',
|
||||
type = 'string', metavar = 'string',
|
||||
help = 'column heading for weights')
|
||||
parser.add_option('--range',
|
||||
dest = 'range',
|
||||
type = 'float', nargs = 2, metavar = 'float float',
|
||||
help = 'data range of histogram [min - max]')
|
||||
parser.add_option('-N',
|
||||
dest = 'N',
|
||||
type = 'int', metavar = 'int',
|
||||
help = 'number of bins')
|
||||
parser.add_option('--density',
|
||||
dest = 'density',
|
||||
action = 'store_true',
|
||||
help = 'report probability density')
|
||||
parser.add_option('--logarithmic',
|
||||
dest = 'log',
|
||||
action = 'store_true',
|
||||
help = 'logarithmically spaced bins')
|
||||
parser.set_defaults(data = None,
|
||||
weights = None,
|
||||
range = None,
|
||||
N = None,
|
||||
density = False,
|
||||
log = False,
|
||||
)
|
||||
|
||||
(options,filenames) = parser.parse_args()
|
||||
|
||||
if not options.data: parser.error('no data specified.')
|
||||
if not options.N: parser.error('no bin number specified.')
|
||||
|
||||
if options.log:
|
||||
def forward(x):
|
||||
return np.log(x)
|
||||
def reverse(x):
|
||||
return np.exp(x)
|
||||
else:
|
||||
def forward(x):
|
||||
return x
|
||||
def reverse(x):
|
||||
return x
|
||||
|
||||
|
||||
# --- loop over input files ------------------------------------------------------------------------
|
||||
|
||||
if filenames == []: filenames = [None]
|
||||
|
||||
for name in filenames:
|
||||
try:
|
||||
table = damask.ASCIItable(name = name, readonly = True)
|
||||
except IOError:
|
||||
continue
|
||||
damask.util.report(scriptName,name)
|
||||
|
||||
# ------------------------------------------ read header ------------------------------------------
|
||||
|
||||
table.head_read()
|
||||
|
||||
# ------------------------------------------ sanity checks ----------------------------------------
|
||||
|
||||
errors = []
|
||||
remarks = []
|
||||
|
||||
if table.label_dimension(options.data) != 1: errors.append('data {} are not scalar.'.format(options.data))
|
||||
if options.weights and \
|
||||
table.label_dimension(options.data) != 1: errors.append('weights {} are not scalar.'.format(options.weights))
|
||||
|
||||
if remarks != []: damask.util.croak(remarks)
|
||||
if errors != []:
|
||||
damask.util.croak(errors)
|
||||
table.close(dismiss = True)
|
||||
continue
|
||||
|
||||
# --------------- read data ----------------------------------------------------------------
|
||||
|
||||
table.data_readArray([options.data,options.weights])
|
||||
|
||||
# --------------- auto range ---------------------------------------------------------------
|
||||
|
||||
if options.range is None:
|
||||
rangeMin,rangeMax = min(table.data[:,0]),max(table.data[:,0])
|
||||
else:
|
||||
rangeMin,rangeMax = min(options.range),max(options.range)
|
||||
|
||||
# --------------- bin data ----------------------------------------------------------------
|
||||
|
||||
count,edges = np.histogram(table.data[:,0],
|
||||
bins = reverse(forward(rangeMin) + np.arange(options.N+1) *
|
||||
(forward(rangeMax)-forward(rangeMin))/options.N),
|
||||
range = (rangeMin,rangeMax),
|
||||
weights = None if options.weights is None else table.data[:,1],
|
||||
density = options.density,
|
||||
)
|
||||
bincenter = reverse(forward(rangeMin) + (0.5+np.arange(options.N)) *
|
||||
(forward(rangeMax)-forward(rangeMin))/options.N) # determine center of bins
|
||||
|
||||
# ------------------------------------------ assemble header ---------------------------------------
|
||||
|
||||
table.info_clear()
|
||||
table.info_append([scriptID + '\t' + ' '.join(sys.argv[1:]),
|
||||
scriptID + ':\t' +
|
||||
'data range {} -- {}'.format(rangeMin,rangeMax) +
|
||||
(' (log)' if options.log else ''),
|
||||
])
|
||||
table.labels_clear()
|
||||
table.labels_append(['bincenter','count'])
|
||||
table.head_write()
|
||||
|
||||
# ------------------------------------------ output result -----------------------------------------
|
||||
|
||||
table.data = np.squeeze(np.dstack((bincenter,count)))
|
||||
table.data_writeArray()
|
||||
|
||||
# ------------------------------------------ output finalization -----------------------------------
|
||||
|
||||
table.close() # close ASCII tables
|
File diff suppressed because it is too large
Load Diff
Loading…
Reference in New Issue