Mercurial > pylearn

comparison pmat.py @ 380:c2f17f231960

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
added function to load amat file
author Frederic Bastien <bastienf@iro.umontreal.ca>
date Wed, 09 Jul 2008 16:55:27 -0400
parents 9330d941fa1f
children
comparison
equal deleted inserted replaced
379:74b402b5a81b 380:c2f17f231960
34 34
35 # Author: Pascal Vincent 35 # Author: Pascal Vincent
36 36
37 #import numarray, sys, os, os.path 37 #import numarray, sys, os, os.path
38 import numpy.numarray, sys, os, os.path 38 import numpy.numarray, sys, os, os.path
39 import fpconst
39 40
40 def array_columns( a, cols ): 41 def array_columns( a, cols ):
41 indices = None 42 indices = None
42 if isinstance( cols, int ): 43 if isinstance( cols, int ):
43 indices = [ cols ] 44 indices = [ cols ]
91 row = row.split() 92 row = row.split()
92 if len(row)>0: 93 if len(row)>0:
93 fieldnames.append(row[0]) 94 fieldnames.append(row[0])
94 f.close() 95 f.close()
95 else: 96 else:
97 self.fieldnames = [ "field_"+str(i) for i in range(a.shape[1]) ]
98
99 return dataset.ArrayDataSet(a,lookup_list.LookupList(fieldnames,[x for x in range(a.shape[1])]))
100
101 def load_amat_as_array_dataset(fname):
102 import dataset,lookup_list
103
104 #load the amat as array
105 (a,fieldnames)=readAMat(fname)
106
107 #load the fieldnames
108 if len(fieldnames)==0:
96 self.fieldnames = [ "field_"+str(i) for i in range(a.shape[1]) ] 109 self.fieldnames = [ "field_"+str(i) for i in range(a.shape[1]) ]
97 110
98 return dataset.ArrayDataSet(a,lookup_list.LookupList(fieldnames,[x for x in range(a.shape[1])])) 111 return dataset.ArrayDataSet(a,lookup_list.LookupList(fieldnames,[x for x in range(a.shape[1])]))
99 112
100 def save_array_dataset_as_pmat(fname,ds): 113 def save_array_dataset_as_pmat(fname,ds):
438 self.putRow(i,row) 451 self.putRow(i,row)
439 452
440 def __len__(self): 453 def __len__(self):
441 return self.length 454 return self.length
442 455
456
457
458 #copied from PLEARNDIR:python_modules/plearn/vmat/readAMat.py
459 def safefloat(str):
460 """Convert the given string to its float value. It is 'safe' in the sense
461 that missing values ('nan') will be properly converted to the corresponding
462 float value under all platforms, contrarily to 'float(str)'.
463 """
464 if str.lower() == 'nan':
465 return fpconst.NaN
466 else:
467 return float(str)
468
469 #copied from PLEARNDIR:python_modules/plearn/vmat/readAMat.py
470 def readAMat(amatname):
471 """Read a PLearn .amat file and return it as a numarray Array.
472
473 Return a tuple, with as the first argument the array itself, and as
474 the second argument the fieldnames (list of strings).
475 """
476 ### NOTE: this version is much faster than first creating the array and
477 ### updating each row as it is read... Bizarrely enough
478 f = open(amatname)
479 a = []
480 fieldnames = []
481 for line in f:
482 if line.startswith("#size:"):
483 (length,width) = line[6:].strip().split()
484 elif line.startswith("#sizes:"): # ignore input/target/weight/extra sizes
485 continue
486
487 elif line.startswith("#:"):
488 fieldnames = line[2:].strip().split()
489 pass
490 elif not line.startswith('#'):
491 # Add all non-comment lines.
492 row = [ safefloat(x) for x in line.strip().split() ]
493 if row:
494 a.append(row)
495
496 f.close()
497 return numpy.numarray.array(a), fieldnames
498
443 499
444 if __name__ == '__main__': 500 if __name__ == '__main__':
445 pmat = PMat( 'tmp.pmat', 'w', fieldnames=['F1', 'F2'] ) 501 pmat = PMat( 'tmp.pmat', 'w', fieldnames=['F1', 'F2'] )
446 pmat.append( [1, 2] ) 502 pmat.append( [1, 2] )
447 pmat.append( [3, 4] ) 503 pmat.append( [3, 4] )