This program measures the time needed to read an array from an hdf file and save it in a numpy array using a slow asarray() method and a fast method with the dtype provided. Run the program as $ python timeNumpyArrayConversion.py <tempfilename.h5> 100 100

timeNumpyArrayConversion.py

#################################################################################
#                                                                               #
# This program measures the time needed to read an array from an hdf file and   #
# save it in a numpy array using a slow asarray() method and a fast method with #
# the dtype explicitly provided before construction. The fast method is to be   #
# preferred for converting arrays read from hdf files using PyTables to Numpy   #
# arrays.		                                              		#			
#			                                                     	#		                                                                                  
# This file is meant to be run as in the following example:	            	#                                                                               
#			                                                     	#
#  $ python timeNumpyArrayConversion.py <tempfilename.h5> <xdim> <ydim> 	#
#  $ python timeNumpyArrayConversion.py myTempfile.h5 1000 1000          	#
#			                                                     	#		    
#################################################################################

import sys, time, h5py, numpy


filename=sys.argv[1] # read in the filename that will be created for the test
x=float(sys.argv[2]) # the x dimension of the array to be created 
y=float(sys.argv[3]) # the y dimension

shape=(x,y)


print "The shape of the array is ", shape
print ""

#open an hdf5 file
f = h5py.File(filename,'w')
grp = f.create_group('grp1')

#create a dataset
datasetWrite = grp.create_dataset('datasetWrite',shape,'=f8',maxshape=(None,None))
f['datasetWrite']=numpy.array(numpy.arange(x*y)).reshape(shape)

#close the file
f.close()

#########################################################################################
# Checking time to read an array from an hdf5 file and converting it into a numpy array #
#########################################################################################

def slow_conversion(hdfarray):
	return numpy.array(hdfarray)
	
def fast_conversion(hdfarray):
	a=numpy.empty(shape=hdfarray.shape,dtype=hdfarray.dtype)
	a[:]=hdfarray[:]
	return a

# slow conversion from a pytables array to a numpy array
t1=time.time()
hf=h5py.File(filename,'r')
numpyArray_Slow = slow_conversion(f['datasetWrite'])
print "Slow conversion time: %8.4f s" %(time.time() - t1) 
hf.close()

# fast conversion from a pytables array to a numpy array
t2=time.time()
hf=h5py.File(filename,'r')
numpyArray_Fast = fast_conversion(f['datasetWrite'])
print "Fast conversion time: %8.4f s" %(time.time() - t2) 
hf.close()