"""Small but quite comprehensive example showing the use of PyTables. The program creates an output file, 'tutorial1.h5'. You can view it with any HDF5 generic utility. """ import os, traceback SECTION = "I HAVE NO TITLE" def tutsep(): print '----8<----', SECTION, '----8<----' def tutprint(obj): tutsep() print obj def tutrepr(obj): tutsep() print repr(obj) def tutexc(): tutsep() traceback.print_exc(file=sys.stdout) SECTION = "Importing tables objects" from numpy import * from tables import * SECTION = "Declaring a Column Descriptor" # Define a user record to characterize some kind of particles class Particle(IsDescription): name = StringCol(16) # 16-character String idnumber = Int64Col() # Signed 64-bit integer ADCcount = UInt16Col() # Unsigned short integer TDCcount = UInt8Col() # unsigned byte grid_i = Int32Col() # integer grid_j = IntCol() # integer (equivalent to Int32Col) pressure = Float32Col() # float (single-precision) energy = FloatCol() # double (double-precision) SECTION = "Creating a PyTables file from scratch" # Open a file in "w"rite mode h5file = openFile('tutorial1.h5', mode = "w", title = "Test file") SECTION = "Creating a new group" # Create a new group under "/" (root) group = h5file.createGroup("/", 'detector', 'Detector information') SECTION = "Creating a new table" # Create one table on it table = h5file.createTable(group, 'readout', Particle, "Readout example") tutprint(h5file) tutrepr(h5file) # Get a shortcut to the record object in table particle = table.row # Fill the table with 10 particles for i in xrange(10): particle['name'] = 'Particle: %6d' % (i) particle['TDCcount'] = i % 256 particle['ADCcount'] = (i * 256) % (1 << 16) particle['grid_i'] = i particle['grid_j'] = 10 - i particle['pressure'] = float(i*i) particle['energy'] = float(particle['pressure'] ** 4) particle['idnumber'] = i * (2 ** 34) # Insert a new particle record particle.append() # Flush the buffers for table table.flush() SECTION = "Reading (and selecting) data in a table" # Read actual data from table. We are interested in collecting pressure values # on entries where TDCcount field is greater than 3 and pressure less than 50 table = h5file.root.detector.readout pressure = [ x['pressure'] for x in table if x['TDCcount']>3 and 20<=x['pressure']<50 ] tutrepr(pressure) # Read also the names with the same cuts names = [ x['name'] for x in table if x['TDCcount'] > 3 and 20 <= x['pressure'] < 50 ] tutrepr(names) SECTION = "Creating new array objects" gcolumns = h5file.createGroup(h5file.root, "columns", "Pressure and Name") tutrepr( h5file.createArray(gcolumns, 'pressure', array(pressure), "Pressure column selection") ) tutrepr( h5file.createArray('/columns', 'name', names, "Name column selection") ) tutprint(h5file) SECTION = "Closing the file and looking at its content" # Close the file h5file.close() tutsep() os.system('h5ls -rd tutorial1.h5') tutsep() os.system('ptdump tutorial1.h5') """This example shows how to browse the object tree and enlarge tables. Before to run this program you need to execute first tutorial1-1.py that create the tutorial1.h5 file needed here. """ SECTION = "Traversing the object tree" # Reopen the file in append mode h5file = openFile("tutorial1.h5", "a") # Print the object tree created from this filename # List all the nodes (Group and Leaf objects) on tree tutprint(h5file) # List all the nodes (using File iterator) on tree tutsep() for node in h5file: print node # Now, only list all the groups on tree tutsep() for group in h5file.walkGroups("/"): print group # List only the arrays hanging from / tutsep() for group in h5file.walkGroups("/"): for array in h5file.listNodes(group, classname = 'Array'): print array # This gives the same result tutsep() for array in h5file.walkNodes("/", "Array"): print array # And finally, list only leafs on /detector group (there should be one!) # Other way using iterators and natural naming tutsep() for leaf in h5file.root.detector('Leaf'): print leaf SECTION = "Setting and getting user attributes" # Get a pointer to '/detector/readout' table = h5file.root.detector.readout # Attach it a string (date) attribute table.attrs.gath_date = "Wed, 06/12/2003 18:33" # Attach a floating point attribute table.attrs.temperature = 18.4 table.attrs.temp_scale = "Celsius" # Get a pointer to '/detector' detector = h5file.root.detector # Attach a general object to the parent (/detector) group detector._v_attrs.stuff = [5, (2.3, 4.5), "Integer and tuple"] # Now, get the attributes tutrepr(table.attrs.gath_date) tutrepr(table.attrs.temperature) tutrepr(table.attrs.temp_scale) tutrepr(detector._v_attrs.stuff) # Delete permanently the attribute gath_date of /detector/readout del table.attrs.gath_date # Print a representation of all attributes in /detector/table tutrepr(table.attrs) # Get the (user) attributes of /detector/table tutprint(table.attrs._f_list("user")) # Get the (sys) attributes of /detector/table tutprint(table.attrs._f_list("sys")) # Rename an attribute table.attrs._f_rename("temp_scale", "tempScale") tutprint(table.attrs._f_list()) # Try to rename a system attribute: try: table.attrs._f_rename("VERSION", "version") except: tutexc() h5file.flush() tutsep() os.system('h5ls -vr tutorial1.h5/detector/readout') SECTION = "Getting object metadata" # Get metadata from table tutsep() print "Object:", table tutsep() print "Table name:", table.name tutsep() print "Table title:", table.title tutsep() print "Number of rows in table:", table.nrows tutsep() print "Table variable names with their type and shape:" tutsep() for name in table.colnames: print name, ':= %s, %s' % (table.coltypes[name], table.colshapes[name]) tutprint(table.__doc__) # Get the object in "/columns pressure" pressureObject = h5file.getNode("/columns", "pressure") # Get some metadata on this object tutsep() print "Info on the object:", repr(pressureObject) tutsep() print " shape: ==>", pressureObject.shape tutsep() print " title: ==>", pressureObject.title tutsep() print " type: ==>", pressureObject.type SECTION = "Reading data from Array objects" # Read the 'pressure' actual data pressureArray = pressureObject.read() tutrepr(pressureArray) tutsep() print "pressureArray is an object of type:", type(pressureArray) # Read the 'name' Array actual data nameArray = h5file.root.columns.name.read() tutrepr(nameArray) print "nameArray is an object of type:", type(nameArray) # Print the data for both arrays tutprint("Data on arrays nameArray and pressureArray:") tutsep() for i in range(pressureObject.shape[0]): print nameArray[i], "-->", pressureArray[i] tutrepr(pressureObject.name) SECTION = "Appending data to an existing table" # Create a shortcut to table object table = h5file.root.detector.readout # Get the object row from table particle = table.row # Append 5 new particles to table for i in xrange(10, 15): particle['name'] = 'Particle: %6d' % (i) particle['TDCcount'] = i % 256 particle['ADCcount'] = (i * 256) % (1 << 16) particle['grid_i'] = i particle['grid_j'] = 10 - i particle['pressure'] = float(i*i) particle['energy'] = float(particle['pressure'] ** 4) particle['idnumber'] = i * (2 ** 34) # This exceeds long integer range particle.append() # Flush this table table.flush() # Print the data using the table iterator: tutsep() for r in table: print "%-16s | %11.1f | %11.4g | %6d | %6d | %8d |" % \ (r['name'], r['pressure'], r['energy'], r['grid_i'], r['grid_j'], r['TDCcount']) # Delete some rows on the Table (yes, rows can be removed!) tutrepr(table.removeRows(5, 10)) # Close the file h5file.close()