def print_bufr_data_category(input_bufr_file):
# #[
"""
an example routine to demonstrate how to retrieve
some meta datafrom the BUFR messages in a BUFR file
"""
# get an instance of the RawBUFRFile class
rbf = RawBUFRFile()
# open the file for reading, count nr of BUFR messages in it
# and store its content in memory, together with
# an array of pointers to the start and end of each BUFR message
rbf.open(input_bufr_file, 'rb')
# extract the number of BUFR messages from the file
num_msgs = rbf.get_num_bufr_msgs()
for msg_nr in range(1, num_msgs + 1):
raw_msg = rbf.get_raw_bufr_msg(msg_nr)[0]
bufr_obj = BUFRInterfaceECMWF(encoded_message=raw_msg)
bufr_obj.decode_sections_012()
data_category = bufr_obj.ksec1[11 - 1]
print('BUFR msg %i has data category %i' % (msg_nr, data_category))
# close the file
rbf.close()
def print_bufr_edition_number(input_bufr_file):
# #[
"""
an example routine to demonstrate how to retrieve
some meta data from the BUFR messages in a BUFR file
"""
# get an instance of the RawBUFRFile class
rbf = RawBUFRFile()
# open the file for reading, count nr of BUFR messages in it
# and store its content in memory, together with
# an array of pointers to the start and end of each BUFR message
rbf.open(input_bufr_file, 'rb')
# extract the number of BUFR messages from the file
num_msgs = rbf.get_num_bufr_msgs()
for msg_nr in range(1, num_msgs+1):
raw_msg = rbf.get_raw_bufr_msg(msg_nr)[0]
bufr_obj = BUFRInterfaceECMWF(encoded_message=raw_msg)
bufr_obj.decode_sections_012()
bufr_edition = bufr_obj.ksec0[3-1]
print('BUFR msg %i has version %i' % (msg_nr, bufr_edition))
# close the file
rbf.close()
def test_get_exp_bufr_table_names(self):
# #[
"""
test the get_expected_ecmwf_bufr_table_names method
"""
center = 210 # = ksec1( 3)
subcenter = 0 # = ksec1(16)
local_version = 1 # = ksec1( 8)
master_table_version = 0 # = ksec1(15)
edition_number = 3 # = ksec0( 3)
master_table_number = 0 # = ksec1(14)
bufrobj = BUFRInterfaceECMWF()
# inspect the location of the ecmwfbufr*.so file, and derive
# from this the location of the BUFR tables that are delivered
# with the ECMWF BUFR library software
ecmwfbufr_path = os.path.split(ecmwfbufr.__file__)[0]
path1 = os.path.join(ecmwfbufr_path, "ecmwf_bufrtables")
path2 = os.path.join(ecmwfbufr_path, '..', "ecmwf_bufrtables")
if os.path.exists(path1):
ecmwf_bufr_tables_dir = path1
elif os.path.exists(path2):
ecmwf_bufr_tables_dir = path2
else:
print("Error: could not find BUFR tables directory")
raise IOError
# make sure the path is absolute, otherwise the ECMWF library
# might fail when it attempts to use it ...
bufrobj.ecmwf_bufr_tables_dir = os.path.abspath(ecmwf_bufr_tables_dir)
(btable, ctable, dtable) = \
bufrobj.get_expected_ecmwf_bufr_table_names(\
center,
subcenter,
local_version,
master_table_version,
edition_number,
master_table_number)
# print("tabel name B: ", btable)
# print("tabel name D: ", dtable)
self.assertEqual(btable, 'B0000000000210000001.TXT')
self.assertEqual(ctable, 'C0000000000210000001.TXT')
self.assertEqual(dtable, 'D0000000000210000001.TXT')
def decoding_example(input_bufr_file, custom_bufr_tables=None):
# #[
"""
wrap the example in a function to circumvent the pylint
convention of requiring capitals for constants in the global
scope (since most of these variables are not constants at all))
"""
# read the binary data using the BUFRFile class
print('loading testfile: ', input_bufr_file)
rbf = RawBUFRFile(verbose=False)
rbf.open(input_bufr_file, 'rb')
(words, section_sizes, section_start_locations) = \
rbf.get_next_raw_bufr_msg()
rbf.close()
if words is None:
print('No valid BUFR messages found')
sys.exit(0)
print('------------------------------')
bufr = BUFRInterfaceECMWF(encoded_message=words,
section_sizes=section_sizes,
section_start_locations=section_start_locations,
verbose=True)
print("calling: decode_sections_012():")
bufr.decode_sections_012()
print("Metadata for decoded BUFR message:")
bufr.print_sections_012_metadata()
print("calling: setup_tables()")
if custom_bufr_tables:
bufr.setup_tables(table_b_to_use=custom_bufr_tables[0],
table_c_to_use=custom_bufr_tables[1],
table_d_to_use=custom_bufr_tables[2])
else:
bufr.setup_tables()
print("calling: print_sections_012():")
bufr.print_sections_012()
# seems not to work correctly now ...
#bufr.fill_descriptor_list()
#bufr.print_descriptors()
print('------------------------------')
print("calling: bufr.decode_data():")
bufr.decode_data()
bufr.decode_sections_0123()
bufr.fill_descriptor_list_subset(subset=1)
return bufr
def encoding_example(output_bufr_file):
# #[
"""
wrap the example in a function to circumvent the pylint
convention of requiring capitals for constants in the global
scope (since most of these variables are not constants at all))
"""
bufr = BUFRInterfaceECMWF(verbose=True)
# fill sections 0, 1, 2 and 3
num_subsets = 4
bufr.fill_sections_0123(bufr_code_centre=98, # ECMWF
bufr_obstype=3, # sounding
bufr_subtype=251, # L1B
bufr_table_local_version=1,
bufr_table_master=0,
bufr_table_master_version=15,
bufr_code_subcentre=0, # L2B processing facility
num_subsets=num_subsets,
bufr_compression_flag=0)
# 64=compression/0=no compression
# determine information from sections 0123 to construct the BUFR table
# names expected by the ECMWF BUFR library and create symlinks to the
# default tables if needed
# NOTE: these custom BUFR tables have been created by the
# create_bufr_tables.py example script
bufr.setup_tables(table_b_to_use='B_my_test_BUFR_table.txt',
table_c_to_use='C_my_test_BUFR_table.txt',
table_d_to_use='D_my_test_BUFR_table.txt')
# define a descriptor list
template = BufrTemplate(verbose=True)
template.add_descriptors(DD_B_048001, # 0
DD_B_048002, # 1
DD_D_348001) # 2
bufr.register_and_expand_descriptors(template)
# activate this one if the encoding crashes without clear cause:
# bufr.estimated_num_bytes_for_encoding = 25000
# retrieve the length of the expanded descriptor list
exp_descr_list_length = bufr.ktdexl
print("exp_descr_list_length = ", exp_descr_list_length)
# fill the values array with some dummy varying data
num_values = exp_descr_list_length*num_subsets
values = np.zeros(num_values, dtype=np.float64) # this is the default
# note: these two must be identical for now, otherwise the
# python to fortran interface breaks down. This also ofcourse is the
# cause of the huge memory use of cvals in case num_values is large.
num_cvalues = num_values
cvals = np.zeros((num_cvalues, 80), dtype=np.character)
for subset in range(num_subsets):
# note that python starts counting with 0, unlike fortran,
# so there is no need to take (subset-1)
i = subset*exp_descr_list_length
# fill the message with some dummy data
values[i] = 1.2515 + 0.0011*subset
i = i+1
values[i] = (3.4562 + 0.0012*subset)*1.e-9
i = i+1
values[i] = 1.2625 + 0.0003*subset
i = i+1
values[i] = (3.4561 + 0.0014*subset)*1.e-9
# do the encoding to binary format
bufr.encode_data(values, cvals)
# get an instance of the RawBUFRFile class
bf1 = RawBUFRFile()
# open the file for writing
bf1.open(output_bufr_file, 'wb')
# write the encoded BUFR message
bf1.write_raw_bufr_msg(bufr.encoded_message)
# close the file
bf1.close()
# define the table name without preceding 'B' or 'D' character
# (which will be prepended by the below write method)
table_name = '_test_table.txt'
bufr_table_set.write_tables(table_name)
# now use these definitions to create a BUFR template
max_nr_of_repeats = 5
num_subsets = 2
num_repetitions = [3, 5] # actual number to be used
template = BufrTemplate(verbose=True)
template.add_descriptor(D_363192) # 1 item
template.del_repl_max_nr_of_repeats_list = [max_nr_of_repeats,]*num_subsets
# and use this BUFR template to create a test BUFR message
bufr = BUFRInterfaceECMWF(verbose=True)
# fill sections 0, 1, 2 and 3
bufr.fill_sections_0123(bufr_code_centre=0,
bufr_obstype=0,
bufr_subtype=0,
bufr_table_local_version=0,
bufr_table_master=0,
bufr_table_master_version=0,
bufr_code_subcentre=0,
num_subsets=num_subsets,
bufr_compression_flag=0)
# 64=compression/0=no compression
# determine information from sections 0123 to construct the BUFR table
# names expected by the ECMWF BUFR library and create symlinks to the
def encoding_example(output_bufr_file):
# #[
"""
wrap the example in a function to circumvent the pylint
convention of requiring capitals for constants in the global
scope (since most of these variables are not constants at all))
"""
bufr = BUFRInterfaceECMWF(verbose=True)
# fill sections 0, 1, 2 and 3
num_subsets = 4
bufr.fill_sections_0123(bufr_code_centre=98, # ECMWF
bufr_obstype=3, # sounding
bufr_subtype=251, # L1B
bufr_table_local_version=1,
bufr_table_master=0,
bufr_table_master_version=15,
bufr_code_subcentre=0, # L2B processing facility
num_subsets=num_subsets,
bufr_compression_flag=0)
# 64=compression/0=no compression
# determine information from sections 0123 to construct the BUFR table
# names expected by the ECMWF BUFR library and create symlinks to the
# default tables if needed
bufr.setup_tables()
# define a descriptor list
template = BufrTemplate()
template.add_descriptors(DD_D_DATE_YYYYMMDD, # 0
DD_D_TIME_HHMM) # 1
# delay replication for the next 2 descriptors
# allow at most 2 delayed replications
template.add_delayed_replic_descriptors(2,
DD_PRESSURE,
DD_TEMPERATURE)
# replicate the next 2 descriptors 3 times
template.add_replicated_descriptors(3,
DD_LATITUDE_HIGH_ACCURACY,
DD_LONGITUDE_HIGH_ACCURACY)
bufr.register_and_expand_descriptors(template)
# activate this one if the encoding crashes without clear cause:
# bufr.estimated_num_bytes_for_encoding = 25000
# retrieve the length of the expanded descriptor list
exp_descr_list_length = bufr.ktdexl
print "exp_descr_list_length = ", exp_descr_list_length
# fill the values array with some dummy varying data
num_values = exp_descr_list_length*num_subsets
values = np.zeros(num_values, dtype=np.float64) # this is the default
# note: these two must be identical for now, otherwise the
# python to fortran interface breaks down. This also ofcourse is the
# cause of the huge memory use of cvals in case num_values is large.
num_cvalues = num_values
cvals = np.zeros((num_cvalues, 80), dtype=np.character)
for subset in range(num_subsets):
# note that python starts counting with 0, unlike fortran,
# so there is no need to take (subset-1)
i = subset*exp_descr_list_length
values[i] = 1999 # year
i = i+1
values[i] = 12 # month
i = i+1
values[i] = 31 # day
i = i+1
values[i] = 23 # hour
i = i+1
values[i] = 59 - subset # minute
i = i+1
values[i] = 2 # delayed replication factor
# this delayed replication factor determines the actual number
# of values to be stored for this particular subset
# even if it is less then the number given in kdata above !
for repl in range(2):
i = i+1
values[i] = 1013.e2 - 100.e2*subset+i+repl # pressure [pa]
i = i+1
values[i] = 273.15 - 10.*subset+i+repl # temperature [K]
for repl in range(3):
i = i+1
values[i] = 51.82 + 0.05*subset+i+repl # latitude
i = i+1
values[i] = 5.25 + 0.1*subset+i+repl # longitude
# do the encoding to binary format
bufr.encode_data(values, cvals)
# get an instance of the RawBUFRFile class
bf1 = RawBUFRFile()
# open the file for writing
bf1.open(output_bufr_file, 'wb')
# write the encoded BUFR message
bf1.write_raw_bufr_msg(bufr.encoded_message)
# close the file
bf1.close()
def create_bufr_file(output_bufr_file, template):
# #[ create bufr file
'''
and use this BUFR template to create a test BUFR message
'''
from pybufr_ecmwf.bufr_interface_ecmwf import BUFRInterfaceECMWF
bufr = BUFRInterfaceECMWF(verbose=True)
# fill sections 0, 1, 2 and 3
num_subsets = 2
bufr.fill_sections_0123(bufr_code_centre=98, # ECMWF
bufr_obstype=3, # sounding
bufr_subtype=253, # L2B
bufr_table_local_version=1,
bufr_table_master=0,
bufr_table_master_version=15,
bufr_code_subcentre=0, # L2B processing facility
num_subsets=num_subsets,
bufr_compression_flag=0)
# 64=compression/0=no compression
# determine information from sections 0123 to construct the BUFR table
# names expected by the ECMWF BUFR library and create symlinks to the
# default tables if needed
bufr.setup_tables(table_b_to_use='B'+TABLE_NAME,
table_d_to_use='D'+TABLE_NAME)
bufr.register_and_expand_descriptors(template)
# activate this one if the encoding crashes without clear cause:
# bufr.estimated_num_bytes_for_encoding = 25000
# retrieve the length of the expanded descriptor list
exp_descr_list_length = bufr.ktdexl
print "exp_descr_list_length = ", exp_descr_list_length
# fill the values array with some dummy varying data
num_values = num_subsets*bufr.max_nr_expanded_descriptors
values = numpy.zeros(num_values, dtype=numpy.float64) # this is the default
# note: these two must be identical for now, otherwise the
# python to fortran interface breaks down. This also ofcourse is the
# cause of the huge memory use of cvals in case num_values is large.
num_cvalues = num_values
cvals = numpy.zeros((num_cvalues, 80), dtype=numpy.character)
cvals_index = 0
repl_counts = []
for subset in range(num_subsets):
# note that python starts counting with 0, unlike fortran,
# so there is no need to take (subset-1)
print 'subset,exp_descr_list_length = ', subset, exp_descr_list_length
i = subset*exp_descr_list_length
# fill the message with some dummy data
# fill year, month, day
for val in [2014, 3, 19]: # fill the header
values[i] = val+subset
i += 1
# fill prod_name
# this is not python2.6 compatible
#txt = 'filename{}.txt'.format(subset)
txt = 'filename'+str(subset)+'.txt'
cvals[cvals_index, :] = ' '
for icval, cval in enumerate(txt):
cvals[cvals_index, icval] = cval
# values[i] = cvals_index * 1000 + 64 # len(txt)
values[i] = (cvals_index+1) * 1000 + len(txt)
i += 1
cvals_index = cvals_index + 1
for val in [5.1+0.1*subset, 55.2-0.01*subset, 23., 45., 73., 82.]:
bufr.verify_in_range(i, val)
values[i] = val
i += 1
if USE_DELAYED_REPLICATION:
# set actual delayed replication repeats
num_repl = 3 + 2*subset
print 'num_repl = ', num_repl
values[i] = num_repl
i += 1
repl_counts.append(num_repl)
# fill the replicated variable
for irepl in range(num_repl):
val = 12.+subset*0.1 + irepl*0.01
bufr.verify_in_range(i, val)
values[i] = val
i += 1
# do the encoding to binary format
bufr.kdata = numpy.array(repl_counts)
print 'bufr.kdata = ', bufr.kdata
bufr.encode_data(values, cvals)
print 'DEBUG: values = ', values
from pybufr_ecmwf.raw_bufr_file import RawBUFRFile
# get an instance of the RawBUFRFile class
bf1 = RawBUFRFile()
# open the file for writing
bf1.open(output_bufr_file, 'wb')
# write the encoded BUFR message
bf1.write_raw_bufr_msg(bufr.encoded_message)
# close the file
bf1.close()
num_replications1 = [1, 2, 3]
num_replications2 = [[
1,
], [
2,
2,
], [3, 3, 3]]
template = BufrTemplate(verbose=True)
template.add_descriptor(D_363192) # 1 item
template.del_repl_max_nr_of_repeats_list = ([
max_nr_of_replications,
] * max_nr_of_replications * num_subsets)
# and use this BUFR template to create a test BUFR message
bufr = BUFRInterfaceECMWF(verbose=True)
# fill sections 0, 1, 2 and 3
bufr.fill_sections_0123(bufr_code_centre=0,
bufr_obstype=0,
bufr_subtype=0,
bufr_table_local_version=0,
bufr_table_master=0,
bufr_table_master_version=0,
bufr_code_subcentre=0,
num_subsets=num_subsets,
bufr_compression_flag=0)
# 64=compression/0=no compression
# determine information from sections 0123 to construct the BUFR table
# names expected by the ECMWF BUFR library and create symlinks to the
def print_bufr_content3(input_bufr_file, output_fd, separator, max_msg_nr,
expand_flags):
# #[ implementation 3
"""
example implementation using the BUFRInterfaceECMWF class
"""
if expand_flags:
print('Sorry, expand_flags is not yet implemented ' +
'for example implementation 3')
# get an instance of the RawBUFRFile class
rbf = RawBUFRFile()
# open the file for reading, count nr of BUFR messages in it
# and store its content in memory, together with
# an array of pointers to the start and end of each BUFR message
rbf.open(input_bufr_file, 'rb')
# extract the number of BUFR messages from the file
num_msgs = rbf.get_num_bufr_msgs()
# print('num_msgs = ',num_msgs)
for msg_nr in range(1, num_msgs + 1):
encoded_message, section_sizes, section_start_locations = \
rbf.get_raw_bufr_msg(msg_nr)
bufr_obj = BUFRInterfaceECMWF(encoded_message, section_sizes,
section_start_locations)
# verbose=True)
bufr_obj.decode_sections_012()
bufr_obj.setup_tables()
# print('num_subsets: ', bufr_obj.get_num_subsets())
# print('num_elements: ',bufr_obj.get_num_elements())
# bufr_obj.decode_sections_0123()
# bufr_obj.print_sections_0123_metadata()
# d = '/home/jos/werk/pybufr_ecmwf_interface/'+\
# 'BUFR_test_files/radar/bufrtables/'
# bufr_obj.setup_tables(table_b_to_use = d+'B0000000000085011012.TXT',
# table_d_to_use = d+'D0000000000085011012.TXT')
# bufr_obj.print_sections_012()
# bufr_obj.fill_descriptor_list()
# do the actual decoding
bufr_obj.decode_data()
# needed to have the units ready, so autoget_cval will work
bufr_obj.decode_sections_0123()
# Create header lines from variable names and units
if msg_nr == 1:
list_of_names = []
list_of_units = []
for (cname, cunit) in zip(bufr_obj.cnames, bufr_obj.cunits):
# glue the ndarray of characters together to form strings
if python3:
cname_str = ''.join(c.decode() for c in cname).strip()
cunit_str = ''.join(c.decode() for c in cunit).strip()
else:
cname_str = ''.join(cname).strip()
cunit_str = ''.join(cunit).strip()
# cnames is a bit over dimensioned, so check for empty values
if cname_str.strip() == '':
break
# append the strings to the head list and quote them
list_of_names.append('"' + cname_str + '"')
list_of_units.append('"' + cunit_str + '"')
output_fd.write('"subset nr"' + separator)
output_fd.write(separator.join(list_of_names) + '\n')
output_fd.write('""' + separator)
output_fd.write(separator.join(list_of_units) + '\n')
nsubsets = bufr_obj.get_num_subsets()
for subs in range(1, nsubsets + 1):
# needed to have the units ready, so autoget_cval will work
bufr_obj.fill_descriptor_list_subset(subs)
nelements = bufr_obj.get_num_elements()
data_list = []
for descr_nr in range(nelements):
data = bufr_obj.get_value(descr_nr, subs, autoget_cval=True)
data_list.append(data)
output_fd.write(
str(subs) + separator +
separator.join(str(val) for val in data_list) + "\n")
print('converted BUFR msg nr. ', msg_nr)
if (max_msg_nr > 0) and (msg_nr >= max_msg_nr):
print('skipping remainder of this BUFR file')
break
# close the BUFR file
rbf.close()
if num_msgs == 0:
print('no BUFR messages found, are you sure this is a BUFR file?')
def print_bufr_content3(input_bufr_file, output_fd, separator,
max_msg_nr, expand_flags):
# #[ implementation 3
"""
example implementation using the BUFRInterfaceECMWF class
"""
if expand_flags:
print('Sorry, expand_flags is not yet implemented '+
'for example implementation 3')
# get an instance of the RawBUFRFile class
rbf = RawBUFRFile()
# open the file for reading, count nr of BUFR messages in it
# and store its content in memory, together with
# an array of pointers to the start and end of each BUFR message
rbf.open(input_bufr_file, 'rb')
# extract the number of BUFR messages from the file
num_msgs = rbf.get_num_bufr_msgs()
# print('num_msgs = ',num_msgs)
for msg_nr in range(1, num_msgs+1):
encoded_message, section_sizes, section_start_locations = \
rbf.get_raw_bufr_msg(msg_nr)
bufr_obj = BUFRInterfaceECMWF(encoded_message, section_sizes,
section_start_locations)
# verbose=True)
bufr_obj.decode_sections_012()
bufr_obj.setup_tables()
# print('num_subsets: ', bufr_obj.get_num_subsets())
# print('num_elements: ',bufr_obj.get_num_elements())
# bufr_obj.decode_sections_0123()
# bufr_obj.print_sections_0123_metadata()
# d = '/home/jos/werk/pybufr_ecmwf_interface/'+\
# 'BUFR_test_files/radar/bufrtables/'
# bufr_obj.setup_tables(table_b_to_use = d+'B0000000000085011012.TXT',
# table_d_to_use = d+'D0000000000085011012.TXT')
# bufr_obj.print_sections_012()
# bufr_obj.fill_descriptor_list()
# do the actual decoding
bufr_obj.decode_data()
# needed to have the units ready, so autoget_cval will work
bufr_obj.decode_sections_0123()
# Create header lines from variable names and units
if msg_nr == 1:
list_of_names = []
list_of_units = []
for (cname, cunit) in zip(bufr_obj.cnames, bufr_obj.cunits):
# glue the ndarray of characters together to form strings
if python3:
cname_str = ''.join(c.decode() for c in cname).strip()
cunit_str = ''.join(c.decode() for c in cunit).strip()
else:
cname_str = ''.join(cname).strip()
cunit_str = ''.join(cunit).strip()
# cnames is a bit over dimensioned, so check for empty values
if cname_str.strip() == '':
break
# append the strings to the head list and quote them
list_of_names.append('"'+cname_str+'"')
list_of_units.append('"'+cunit_str+'"')
output_fd.write('"subset nr"'+separator)
output_fd.write(separator.join(list_of_names) + '\n')
output_fd.write('""'+separator)
output_fd.write(separator.join(list_of_units) + '\n')
nsubsets = bufr_obj.get_num_subsets()
for subs in range(1, nsubsets+1):
# needed to have the units ready, so autoget_cval will work
bufr_obj.fill_descriptor_list_subset(subs)
nelements = bufr_obj.get_num_elements()
data_list = []
for descr_nr in range(nelements):
data = bufr_obj.get_value(descr_nr, subs, autoget_cval=True)
data_list.append(data)
output_fd.write(str(subs)+separator+
separator.join(str(val) for val in data_list)+
"\n")
print('converted BUFR msg nr. ', msg_nr)
if (max_msg_nr > 0) and (msg_nr >= max_msg_nr):
print('skipping remainder of this BUFR file')
break
# close the BUFR file
rbf.close()
if num_msgs == 0:
print('no BUFR messages found, are you sure this is a BUFR file?')
def create_bufr_file(output_bufr_file, template):
# #[ create bufr file
'''
and use this BUFR template to create a test BUFR message
'''
from pybufr_ecmwf.bufr_interface_ecmwf import BUFRInterfaceECMWF
bufr = BUFRInterfaceECMWF(verbose=True)
# fill sections 0, 1, 2 and 3
num_subsets = 2
bufr.fill_sections_0123(bufr_code_centre=98, # ECMWF
bufr_obstype=3, # sounding
bufr_subtype=253, # L2B
bufr_table_local_version=1,
bufr_table_master=0,
bufr_table_master_version=15,
bufr_code_subcentre=0, # L2B processing facility
num_subsets=num_subsets,
bufr_compression_flag=0)
# 64=compression/0=no compression
# determine information from sections 0123 to construct the BUFR table
# names expected by the ECMWF BUFR library and create symlinks to the
# default tables if needed
bufr.setup_tables(table_b_to_use='B'+TABLE_NAME,
table_d_to_use='D'+TABLE_NAME)
bufr.register_and_expand_descriptors(template)
# activate this one if the encoding crashes without clear cause:
# bufr.estimated_num_bytes_for_encoding = 25000
# retrieve the length of the expanded descriptor list
exp_descr_list_length = bufr.ktdexl
print "exp_descr_list_length = ", exp_descr_list_length
# fill the values array with some dummy varying data
num_values = num_subsets*bufr.max_nr_expanded_descriptors
values = numpy.zeros(num_values, dtype=numpy.float64) # this is the default
# note: these two must be identical for now, otherwise the
# python to fortran interface breaks down. This also ofcourse is the
# cause of the huge memory use of cvals in case num_values is large.
num_cvalues = num_values
cvals = numpy.zeros((num_cvalues, 80), dtype=numpy.character)
cvals_index = 0
repl_counts = []
for subset in range(num_subsets):
# note that python starts counting with 0, unlike fortran,
# so there is no need to take (subset-1)
print 'subset,exp_descr_list_length = ', subset, exp_descr_list_length
i = subset*exp_descr_list_length
# fill the message with some dummy data
# fill year, month, day
for val in [2014, 3, 19]: # fill the header
values[i] = val+subset
i += 1
# fill prod_name
txt = 'filename{}.txt'.format(subset)
cvals[cvals_index, :] = ' '
for icval, cval in enumerate(txt):
cvals[cvals_index, icval] = cval
# values[i] = cvals_index * 1000 + 64 # len(txt)
values[i] = (cvals_index+1) * 1000 + len(txt)
i += 1
cvals_index = cvals_index + 1
for val in [5.1+0.1*subset, 55.2-0.01*subset, 23., 45., 73., 82.]:
bufr.verify_in_range(i, val)
values[i] = val
i += 1
if USE_DELAYED_REPLICATION:
# set actual delayed replication repeats
num_repl = 3 + 2*subset
print 'num_repl = ', num_repl
values[i] = num_repl
i += 1
repl_counts.append(num_repl)
# fill the replicated variable
for irepl in range(num_repl):
val = 12.+subset*0.1 + irepl*0.01
bufr.verify_in_range(i, val)
values[i] = val
i += 1
# do the encoding to binary format
bufr.kdata = numpy.array(repl_counts)
print 'bufr.kdata = ', bufr.kdata
bufr.encode_data(values, cvals)
print 'DEBUG: values = ', values
from pybufr_ecmwf.raw_bufr_file import RawBUFRFile
# get an instance of the RawBUFRFile class
bf1 = RawBUFRFile()
# open the file for writing
bf1.open(output_bufr_file, 'wb')
# write the encoded BUFR message
bf1.write_raw_bufr_msg(bufr.encoded_message)
# close the file
bf1.close()
def encoding_example(output_bufr_file):
# #[
"""
wrap the example in a function to circumvent the pylint
convention of requiring capitals for constants in the global
scope (since most of these variables are not constants at all))
"""
bufr = BUFRInterfaceECMWF(verbose=True)
# fill sections 0, 1, 2 and 3
num_subsets = 4
bufr.fill_sections_0123(
bufr_code_centre=98, # ECMWF
bufr_obstype=3, # sounding
bufr_subtype=251, # L1B
bufr_table_local_version=1,
bufr_table_master=0,
bufr_table_master_version=15,
bufr_code_subcentre=0, # L2B processing facility
num_subsets=num_subsets,
bufr_compression_flag=0)
# 64=compression/0=no compression
# determine information from sections 0123 to construct the BUFR table
# names expected by the ECMWF BUFR library and create symlinks to the
# default tables if needed
bufr.setup_tables()
# define a descriptor list
template = BufrTemplate(verbose=True)
template.add_descriptors(
DD_D_DATE_YYYYMMDD, # 0
DD_D_TIME_HHMM) # 1
# delay replication for the next 2 descriptors
# allow at most 2 delayed replications
template.add_delayed_replic_descriptors(2, DD_PRESSURE, DD_TEMPERATURE)
# replicate the next 2 descriptors 3 times
template.add_replicated_descriptors(3, DD_LATITUDE_HIGH_ACCURACY,
DD_LONGITUDE_HIGH_ACCURACY)
bufr.register_and_expand_descriptors(template)
# activate this one if the encoding crashes without clear cause:
# bufr.estimated_num_bytes_for_encoding = 25000
# retrieve the length of the expanded descriptor list
exp_descr_list_length = bufr.ktdexl
print("exp_descr_list_length = ", exp_descr_list_length)
# fill the values array with some dummy varying data
num_values = exp_descr_list_length * num_subsets
values = np.zeros(num_values, dtype=np.float64) # this is the default
# note: these two must be identical for now, otherwise the
# python to fortran interface breaks down. This also ofcourse is the
# cause of the huge memory use of cvals in case num_values is large.
num_cvalues = num_values
cvals = np.zeros((num_cvalues, 80), dtype=np.character)
for subset in range(num_subsets):
# note that python starts counting with 0, unlike fortran,
# so there is no need to take (subset-1)
i = subset * exp_descr_list_length
values[i] = 1999 # year
i = i + 1
values[i] = 12 # month
i = i + 1
values[i] = 31 # day
i = i + 1
values[i] = 23 # hour
i = i + 1
values[i] = 59 - subset # minute
i = i + 1
values[i] = 2 # delayed replication factor
# this delayed replication factor determines the actual number
# of values to be stored for this particular subset
# even if it is less then the number given in kdata above !
for repl in range(2):
i = i + 1
values[i] = 1013.e2 - 100.e2 * subset + i + repl # pressure [pa]
i = i + 1
values[i] = 273.15 - 10. * subset + i + repl # temperature [K]
for repl in range(3):
i = i + 1
values[i] = 51.82 + 0.05 * subset + i + repl # latitude
i = i + 1
values[i] = 5.25 + 0.1 * subset + i + repl # longitude
# do the encoding to binary format
bufr.encode_data(values, cvals)
# get an instance of the RawBUFRFile class
bf1 = RawBUFRFile()
# open the file for writing
bf1.open(output_bufr_file, 'wb')
# write the encoded BUFR message
bf1.write_raw_bufr_msg(bufr.encoded_message)
# close the file
bf1.close()
def upgrade_bufr_file(input_bufr_file, output_bufr_file):
# #[
"""
an example routine to demonstrate how to read a BUFR message,
upgrade it's edition number, and write it to an output BUFR file
"""
# get 2 instances of the RawBUFRFile class
rbf_in = RawBUFRFile()
rbf_out = RawBUFRFile()
# open the file for reading, count nr of BUFR messages in it
# and store its content in memory, together with
# an array of pointers to the start and end of each BUFR message
rbf_in.open(input_bufr_file, 'rb')
# open the file for writing
rbf_out.open(output_bufr_file, 'wb')
# extract the number of BUFR messages from the file
num_msgs = rbf_in.get_num_bufr_msgs()
for msg_nr in range(1, num_msgs+1):
raw_msg, section_sizes, section_start_locations = \
rbf_in.get_raw_bufr_msg(msg_nr)
bufr_obj = BUFRInterfaceECMWF(raw_msg, section_sizes,
section_start_locations,
verbose=True)
bufr_obj.decode_sections_012()
bufr_obj.setup_tables()
bufr_obj.decode_data()
nsub = bufr_obj.get_num_subsets()
n_exp_descr = len(bufr_obj.values)/nsub
bufr_obj.fill_descriptor_list(nr_of_expanded_descriptors=n_exp_descr)
bufr_obj.ksec0[3-1] = 4 # set bufr edition to 4
bufr_obj.ktdlst = bufr_obj.get_descriptor_list()
# extract delayed replication factors
delayed_repl_data = bufr_obj.derive_delayed_repl_factors()
# fill the list of replication factors
bufr_obj.fill_delayed_repl_data(delayed_repl_data)
# activate this one if the encoding crashes without clear cause:
# bufr_obj.estimated_num_bytes_for_encoding = 25000
# encode the data
bufr_obj.encode_data(bufr_obj.values, bufr_obj.cvals)
print 'Encode BUFR msg %i' % msg_nr
rbf_out.write_raw_bufr_msg(bufr_obj.encoded_message)
# close the file
rbf_in.close()
rbf_out.close()
def decoding_example(input_bufr_file, custom_bufr_tables=None):
# #[
"""
wrap the example in a function to circumvent the pylint
convention of requiring capitals for constants in the global
scope (since most of these variables are not constants at all))
"""
# read the binary data using the BUFRFile class
print('loading testfile: ', input_bufr_file)
rbf = RawBUFRFile(verbose=False)
rbf.open(input_bufr_file, 'rb')
(words, section_sizes, section_start_locations) = \
rbf.get_next_raw_bufr_msg()
rbf.close()
if words is None:
print('No valid BUFR messages found')
sys.exit(0)
print('------------------------------')
bufr = BUFRInterfaceECMWF(encoded_message=words,
section_sizes=section_sizes,
section_start_locations=section_start_locations,
verbose=True)
print("calling: decode_sections_012():")
bufr.decode_sections_012()
print("Metadata for decoded BUFR message:")
bufr.print_sections_012_metadata()
print("calling: setup_tables()")
if custom_bufr_tables:
bufr.setup_tables(table_b_to_use=custom_bufr_tables[0],
table_c_to_use=custom_bufr_tables[1],
table_d_to_use=custom_bufr_tables[2])
else:
bufr.setup_tables()
print("calling: print_sections_012():")
bufr.print_sections_012()
# seems not to work correctly now ...
#bufr.fill_descriptor_list()
#bufr.print_descriptors()
print('------------------------------')
print("calling: bufr.decode_data():")
bufr.decode_data()
bufr.decode_sections_0123()
bufr.fill_descriptor_list_subset(subset=1)
return bufr
def encode(output_bufr_file, RadarData, WMOID):
# Get the data for the wind profilers
WindComponentData = RadarData.getWindComponents()
GateSpaceWidths = RadarData.getGateSpace()
if len(GateSpaceWidths) > 1:
MaxWidth = 0
for Width in GateSpaceWidths:
if Width > MaxWidth:
MaxWidth = Width
else:
MaxWidth = GateSpaces[0]
print('Using Pulse Mode ' + str(MaxWidth))
for DateTime in WindComponentData:
print('Processing ' + str(DateTime))
print(MaxWidth)
print(WindComponentData[DateTime].keys())
if MaxWidth in WindComponentData[DateTime].keys():
print('Processing high mode only. Pulse ' + str(MaxWidth)
+ 'ns')
WindComponents = WindComponentData[DateTime][MaxWidth]
# Create tge buffer for the hour block of data.
bufr = BUFRInterfaceECMWF(verbose=True)
# fill sections 0, 1, 2 and 3 in the BUFR table
num_subsets = 1
bufr.fill_sections_0123( # ECMWF
# wind profiler . Also know as Message Type (Table A)
# Message sub-type
# L2B processing facility
bufr_code_centre=59,
bufr_obstype=2,
bufr_subtype=7,
bufr_table_local_version=1,
bufr_table_master=0,
bufr_table_master_version=3,
bufr_code_subcentre=0,
num_subsets=num_subsets,
bufr_compression_flag=0,
)
bufr.setup_tables()
# define a descriptor list
template = BufrTemplate()
print('adding {0} descriptors'.format(10))
template.add_descriptors(
WMO_BLOCK_NUM,
WMO_STATION_NUM,
DD_LATITUDE_COARSE_ACCURACY,
DD_LONGITUDE_COARSE_ACCURACY,
STATION_HEIGHT,
YEAR,
MONTH,
MDAY,
HOUR,
MIN,
TIME_SIGNIFICANCE,
TIME_PERIOD,
DD_WIND_SPEED,
DD_WIND_DIR,
DD_PRESSURE,
DD_TEMPERATURE,
RAINFALL_SWE,
DD_RELATIVE_HUMD,
HEIGHT_INCREMENT,
WIND_PROFILER_SUB_MODE_INFO,
HEIGHT_INCREMENT,
HEIGHT_INCREMENT,
HEIGHT_INCREMENT,
)
# delay replication for the next 10 descriptors
template.add_replicated_descriptors(
len(WindComponents),
WIND_PROFILER_MODE_INFO,
WIND_PROFILER_QC_RESULTS,
TOTAL_NUMBER,
WIND_U_COMPONENT,
WIND_V_COMPONENT,
STD_DEV_HORIZONTAL_WIND_SPEED,
TOTAL_NUMBER,
RADAR_BACK_SCATTER,
WIND_W_COMPONENT,
STD_DEV_VERTICAL_SPEED,
)
bufr.register_and_expand_descriptors(template)
# activate this one if the encoding crashes without clear cause:
# bufr.estimated_num_bytes_for_encoding = 25000
# retrieve the length of the expanded descriptor list
exp_descr_list_length = bufr.ktdexl
# fill the values array with some dummy varying data
num_values = exp_descr_list_length
values = np.zeros(num_values, dtype=np.float64) # this is the default
# note: these two must be identical for now, otherwise the
# python to fortran interface breaks down. This also ofcourse is the
# cause of the huge memory use of cvals in case num_values is large.
num_cvalues = num_values
cvals = np.zeros((num_cvalues, 80), dtype=np.character)
# note that python starts counting with 0, unlike fortran,
# so there is no need to take (subset-1)
# i = subset*exp_descr_list_length
values[0] = WMOID[0:2] # WMO Block Number
values[1] = WMOID[2:5] # WMO Station #
values[2] = RadarData.getLatitude() # Latitude
values[3] = RadarData.getLongitude()
values[4] = RadarData.getElev() # Elevation of Station (meters)
values[5] = DateTime.timetuple().tm_year # year
values[6] = DateTime.timetuple().tm_mon # month
values[7] = DateTime.timetuple().tm_mday # day
values[8] = DateTime.timetuple().tm_hour # hour
values[9] = 0 # minute
values[10] = 2 # Time Significance
values[11] = -60 # Time Period
values[12] = 1 # Wind Speed
values[13] = 1 # Wind Dir
values[14] = 1 # Pressure
values[15] = 1 # Temperature
values[16] = .2 # Rainfall
values[17] = 1 # Realative Humidty
GateSpace = int(MaxWidth) * 1.e-9 * 3.e+8 / 2.
values[18] = GateSpace # Height Increment
values[19] = 0 # Wind Profiler Sub Mode
values[20] = GateSpace # Height Increment
values[21] = GateSpace # Height Increment
values[22] = GateSpace # Height Increment
print('Number of gates ' + str(len(WindComponents)))
for i in range(0, len(WindComponents)):
for t in range(0, 10):
# Calulcate the correct index in the BUFR
rec = i * 10 + 23 + t
if WindComponents[i][0] <= 1000 \
and WindComponents[i][1] <= 1000:
WindRadians = math.radians(WindComponents[i][1])
VelocityU = -WindComponents[i][0] \
* math.sin(WindRadians)
VelocityV = -WindComponents[i][0] \
* math.cos(WindRadians)
TotalNumber = WindComponents[i][2]
SnrDB = WindComponents[i][3]
else:
VelocityU = VelocityV = TotalNumbe = SnrDB = -99
values[rec] = float('NaN') # level mode
continue
if rec % 10 == 3:
values[rec] = 1 # level mode
if rec % 10 == 4:
values[rec] = 0 # Quality Control test
if rec % 10 == 5:
values[rec] = TotalNumber # Total Number (with respect to accumlation or average)
if rec % 10 == 6:
values[rec] = VelocityU # U-Component
if rec % 10 == 7:
values[rec] = VelocityV # V-Component
if rec % 10 == 8:
values[rec] = 0.000 # Std Deviation of horizontal wind speed
if rec % 10 == 9:
values[rec] = TotalNumber # Total Number (with respect to accumlation or average)
if rec % 10 == 0:
values[rec] = SnrDB / 100 # Radar Back Scatter (Peak Power) x100
if rec % 10 == 1:
values[rec] = 0.000 # W-Component x100
if rec % 10 == 2:
values[rec] = 0.000 # Std Deviation of vertical wind
# do the encoding to binary format
bufr.encode_data(values, cvals)
HeaderString = '''
287
IUAK01 PANC %02d%02d00
''' \
% (DateTime.timetuple().tm_mday,
DateTime.timetuple().tm_hour)
if not os.path.exists(OutputPath):
os.makedirs(OutputPath)
OutputFile = \
'%s/IUPTO2_%s_%02d%02d00_216234297.bufr.%04d%02d%02d%02d' \
% (
OutputPath,
RadarData.getSiteID().upper(),
DateTime.timetuple().tm_mon,
DateTime.timetuple().tm_mday,
DateTime.timetuple().tm_year,
DateTime.timetuple().tm_mon,
DateTime.timetuple().tm_mday,
DateTime.timetuple().tm_hour,
)
# Remove file if exsists
if os.path.exists(OutputFile):
os.remove(OutputFile)
bf1 = open(OutputFile, 'ab')
bf1.write(HeaderString)
bf1.close()
# get an instance of the RawBUFRFile class
bf1 = RawBUFRFile()
# open the file for writing
bf1.open(OutputFile, 'ab')
# write the encoded BUFR message
bf1.write_raw_bufr_msg(bufr.encoded_message)
# close the file
bf1.close()
# #]
print('succesfully written BUFR encoded data to file: ',
OutputFile)
def encode(output_bufr_file, RadarData, WMOID):
# Get the data for the wind profilers
WindComponentData = RadarData.getWindComponents()
GateSpaceWidths = RadarData.getGateSpace()
if len(GateSpaceWidths) > 1:
MaxWidth = 0
for Width in GateSpaceWidths:
if Width > MaxWidth:
MaxWidth = Width
else:
MaxWidth = GateSpaces[0]
print('Using Pulse Mode ' + str(MaxWidth))
for DateTime in WindComponentData:
print('Processing ' + str(DateTime))
print(MaxWidth)
print(WindComponentData[DateTime].keys())
if MaxWidth in WindComponentData[DateTime].keys():
print('Processing high mode only. Pulse ' + str(MaxWidth) + 'ns')
WindComponents = WindComponentData[DateTime][MaxWidth]
# Create tge buffer for the hour block of data.
bufr = BUFRInterfaceECMWF(verbose=True)
# fill sections 0, 1, 2 and 3 in the BUFR table
num_subsets = 1
bufr.fill_sections_0123( # ECMWF
# wind profiler . Also know as Message Type (Table A)
# Message sub-type
# L2B processing facility
bufr_code_centre=59,
bufr_obstype=2,
bufr_subtype=7,
bufr_table_local_version=1,
bufr_table_master=0,
bufr_table_master_version=3,
bufr_code_subcentre=0,
num_subsets=num_subsets,
bufr_compression_flag=0,
)
bufr.setup_tables()
# define a descriptor list
template = BufrTemplate()
print('adding {0} descriptors'.format(10))
template.add_descriptors(
WMO_BLOCK_NUM,
WMO_STATION_NUM,
DD_LATITUDE_COARSE_ACCURACY,
DD_LONGITUDE_COARSE_ACCURACY,
STATION_HEIGHT,
YEAR,
MONTH,
MDAY,
HOUR,
MIN,
TIME_SIGNIFICANCE,
TIME_PERIOD,
DD_WIND_SPEED,
DD_WIND_DIR,
DD_PRESSURE,
DD_TEMPERATURE,
RAINFALL_SWE,
DD_RELATIVE_HUMD,
HEIGHT_INCREMENT,
WIND_PROFILER_SUB_MODE_INFO,
HEIGHT_INCREMENT,
HEIGHT_INCREMENT,
HEIGHT_INCREMENT,
)
# delay replication for the next 10 descriptors
template.add_replicated_descriptors(
len(WindComponents),
WIND_PROFILER_MODE_INFO,
WIND_PROFILER_QC_RESULTS,
TOTAL_NUMBER,
WIND_U_COMPONENT,
WIND_V_COMPONENT,
STD_DEV_HORIZONTAL_WIND_SPEED,
TOTAL_NUMBER,
RADAR_BACK_SCATTER,
WIND_W_COMPONENT,
STD_DEV_VERTICAL_SPEED,
)
bufr.register_and_expand_descriptors(template)
# activate this one if the encoding crashes without clear cause:
# bufr.estimated_num_bytes_for_encoding = 25000
# retrieve the length of the expanded descriptor list
exp_descr_list_length = bufr.ktdexl
# fill the values array with some dummy varying data
num_values = exp_descr_list_length
values = np.zeros(num_values,
dtype=np.float64) # this is the default
# note: these two must be identical for now, otherwise the
# python to fortran interface breaks down. This also ofcourse is the
# cause of the huge memory use of cvals in case num_values is large.
num_cvalues = num_values
cvals = np.zeros((num_cvalues, 80), dtype=np.character)
# note that python starts counting with 0, unlike fortran,
# so there is no need to take (subset-1)
# i = subset*exp_descr_list_length
values[0] = WMOID[0:2] # WMO Block Number
values[1] = WMOID[2:5] # WMO Station #
values[2] = RadarData.getLatitude() # Latitude
values[3] = RadarData.getLongitude()
values[4] = RadarData.getElev() # Elevation of Station (meters)
values[5] = DateTime.timetuple().tm_year # year
values[6] = DateTime.timetuple().tm_mon # month
values[7] = DateTime.timetuple().tm_mday # day
values[8] = DateTime.timetuple().tm_hour # hour
values[9] = 0 # minute
values[10] = 2 # Time Significance
values[11] = -60 # Time Period
values[12] = 1 # Wind Speed
values[13] = 1 # Wind Dir
values[14] = 1 # Pressure
values[15] = 1 # Temperature
values[16] = .2 # Rainfall
values[17] = 1 # Realative Humidty
GateSpace = int(MaxWidth) * 1.e-9 * 3.e+8 / 2.
values[18] = GateSpace # Height Increment
values[19] = 0 # Wind Profiler Sub Mode
values[20] = GateSpace # Height Increment
values[21] = GateSpace # Height Increment
values[22] = GateSpace # Height Increment
print('Number of gates ' + str(len(WindComponents)))
for i in range(0, len(WindComponents)):
for t in range(0, 10):
# Calulcate the correct index in the BUFR
rec = i * 10 + 23 + t
if WindComponents[i][0] <= 1000 \
and WindComponents[i][1] <= 1000:
WindRadians = math.radians(WindComponents[i][1])
VelocityU = -WindComponents[i][0] \
* math.sin(WindRadians)
VelocityV = -WindComponents[i][0] \
* math.cos(WindRadians)
TotalNumber = WindComponents[i][2]
SnrDB = WindComponents[i][3]
else:
VelocityU = VelocityV = TotalNumbe = SnrDB = -99
values[rec] = float('NaN') # level mode
continue
if rec % 10 == 3:
values[rec] = 1 # level mode
if rec % 10 == 4:
values[rec] = 0 # Quality Control test
if rec % 10 == 5:
values[
rec] = TotalNumber # Total Number (with respect to accumlation or average)
if rec % 10 == 6:
values[rec] = VelocityU # U-Component
if rec % 10 == 7:
values[rec] = VelocityV # V-Component
if rec % 10 == 8:
values[
rec] = 0.000 # Std Deviation of horizontal wind speed
if rec % 10 == 9:
values[
rec] = TotalNumber # Total Number (with respect to accumlation or average)
if rec % 10 == 0:
values[
rec] = SnrDB / 100 # Radar Back Scatter (Peak Power) x100
if rec % 10 == 1:
values[rec] = 0.000 # W-Component x100
if rec % 10 == 2:
values[rec] = 0.000 # Std Deviation of vertical wind
# do the encoding to binary format
bufr.encode_data(values, cvals)
HeaderString = '''
287
IUAK01 PANC %02d%02d00
''' \
% (DateTime.timetuple().tm_mday,
DateTime.timetuple().tm_hour)
if not os.path.exists(OutputPath):
os.makedirs(OutputPath)
OutputFile = \
'%s/IUPTO2_%s_%02d%02d00_216234297.bufr.%04d%02d%02d%02d' \
% (
OutputPath,
RadarData.getSiteID().upper(),
DateTime.timetuple().tm_mon,
DateTime.timetuple().tm_mday,
DateTime.timetuple().tm_year,
DateTime.timetuple().tm_mon,
DateTime.timetuple().tm_mday,
DateTime.timetuple().tm_hour,
)
# Remove file if exsists
if os.path.exists(OutputFile):
os.remove(OutputFile)
bf1 = open(OutputFile, 'ab')
bf1.write(HeaderString)
bf1.close()
# get an instance of the RawBUFRFile class
bf1 = RawBUFRFile()
# open the file for writing
bf1.open(OutputFile, 'ab')
# write the encoded BUFR message
bf1.write_raw_bufr_msg(bufr.encoded_message)
# close the file
bf1.close()
# #]
print('succesfully written BUFR encoded data to file: ',
OutputFile)