UPDATE FOR JSON DATABASE INSTEAD OF CSV
This commit is contained in:
parent
605698ca09
commit
22cc9d1920
@ -1,7 +1,7 @@
|
||||
#!/usr/bin/env python
|
||||
#
|
||||
###############################################################################
|
||||
# Copyright (C) 2016 Cortney T. Buffington, N0MJS <n0mjs@me.com>
|
||||
# Copyright (C) 2016-2019 Cortney T. Buffington, N0MJS <n0mjs@me.com>
|
||||
#
|
||||
# This program is free software; you can redistribute it and/or modify
|
||||
# it under the terms of the GNU General Public License as published by
|
||||
@ -20,16 +20,15 @@
|
||||
|
||||
from __future__ import print_function
|
||||
|
||||
from json import load as jload
|
||||
from os.path import isfile, getmtime
|
||||
from time import time
|
||||
from urllib import URLopener
|
||||
from csv import reader as csv_reader
|
||||
from csv import DictReader as csv_dict_reader
|
||||
from binascii import b2a_hex as ahex
|
||||
|
||||
# Does anybody read this stuff? There's a PEP somewhere that says I should do this.
|
||||
__author__ = 'Cortney T. Buffington, N0MJS'
|
||||
__copyright__ = 'Copyright (c) 2016-2017 Cortney T. Buffington, N0MJS and the K0USY Group'
|
||||
__copyright__ = 'Copyright (c) 2016-2019 Cortney T. Buffington, N0MJS and the K0USY Group'
|
||||
__credits__ = 'Colin Durbridge, G4EML, Steve Zingman, N4IRS; Mike Zingman'
|
||||
__license__ = 'GNU GPLv3'
|
||||
__maintainer__ = 'Cort Buffington, N0MJS'
|
||||
@ -97,45 +96,60 @@ def try_download(_path, _file, _url, _stale,):
|
||||
url.close()
|
||||
return result
|
||||
|
||||
# LEGACY VERSION - MAKES A SIMPLE {INTEGER ID: 'CALLSIGN'} DICTIONARY
|
||||
# SHORT VERSION - MAKES A SIMPLE {INTEGER ID: 'CALLSIGN'} DICTIONARY
|
||||
def mk_id_dict(_path, _file):
|
||||
dict = {}
|
||||
_dict = {}
|
||||
try:
|
||||
with open(_path+_file, 'rU') as _handle:
|
||||
ids = csv_reader(_handle, dialect='excel', delimiter=',')
|
||||
for row in ids:
|
||||
try:
|
||||
dict[int(row[0])] = (row[1])
|
||||
except:
|
||||
pass
|
||||
records = jload(_handle)['results']
|
||||
_handle.close
|
||||
return dict
|
||||
for record in records:
|
||||
_dict[int(record['id'])] = record['callsign'].encode('ascii','ignore')
|
||||
return _dict
|
||||
except IOError:
|
||||
return dict
|
||||
return _dict
|
||||
|
||||
# NEW VERSION - MAKES A FULL DICTIONARY OF INFORMATION BASED ON TYPE OF ALIAS FILE
|
||||
# BASED ON DOWNLOADS FROM DMR-MARC, TGID IS STILL A "SIMPLE" DICTIONARY
|
||||
# LONG VERSION - MAKES A FULL DICTIONARY OF INFORMATION BASED ON TYPE OF ALIAS FILE
|
||||
# BASED ON DOWNLOADS FROM RADIOID.NET
|
||||
def mk_full_id_dict(_path, _file, _type):
|
||||
dict = {}
|
||||
if _type == 'subscriber':
|
||||
fields = SUB_FIELDS
|
||||
elif _type == 'peer':
|
||||
fields = PEER_FIELDS
|
||||
elif _type == 'tgid':
|
||||
fields = TGID_FIELDS
|
||||
_dict = {}
|
||||
try:
|
||||
with open(_path+_file, 'rU') as _handle:
|
||||
ids = csv_dict_reader(_handle, fieldnames=fields, restkey='OTHER', dialect='excel', delimiter=',')
|
||||
for row in ids:
|
||||
for item in row:
|
||||
try:
|
||||
dict[int(row['ID'])] = row
|
||||
except:
|
||||
pass
|
||||
records = jload(_handle)['results']
|
||||
_handle.close
|
||||
return dict
|
||||
if _type == 'peer':
|
||||
for record in records:
|
||||
_dict[int(record['id'])] = {
|
||||
'CALLSIGN': record['callsign'],
|
||||
'CITY': record['city'],
|
||||
'STATE': record['state'],
|
||||
'COUNTRY': record['country'],
|
||||
'FREQ': record['frequency'],
|
||||
'CC': record['color_code'],
|
||||
'OFFSET': record['offset'],
|
||||
'LINKED': record['ts_linked'],
|
||||
'TRUSTEE': record['trustee'],
|
||||
'NETWORK': record['ipsc_network']
|
||||
}
|
||||
elif _type == 'subscriber':
|
||||
for record in records:
|
||||
_dict[int(record['id'])] = {
|
||||
'CALLSIGN': record['callsign'],
|
||||
'NAME': (record['fname'] + ' ' + record['surname']),
|
||||
'CITY': record['city'],
|
||||
'STATE': record['state'],
|
||||
'COUNTRY': record['country']
|
||||
}
|
||||
elif _type == 'tgid':
|
||||
for record in records:
|
||||
_dict[int(record['tgid'])] = {
|
||||
'NAME': record['name'],
|
||||
'ID': record['id']
|
||||
}
|
||||
return _dict
|
||||
except IOError:
|
||||
return dict
|
||||
return _dict
|
||||
|
||||
|
||||
# THESE ARE THE SAME THING FOR LEGACY PURPOSES
|
||||
def get_alias(_id, _dict, *args):
|
||||
@ -169,8 +183,56 @@ def get_info(_id, _dict, *args):
|
||||
else:
|
||||
return _dict[_id]
|
||||
return _id
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
|
||||
'''
|
||||
repeater file: ('callsign', 'city', 'color_code', 'country', 'frequency', 'ipsc_network', 'locator', 'offset', 'state', 'trustee', 'ts_linked')
|
||||
user file: ('callsign', 'city', 'country', 'fname', 'radio_id', 'remarks', 'state', 'surname')
|
||||
'''
|
||||
|
||||
# Try updating peer aliases file
|
||||
result = try_download('/tmp/', 'peers.json', 'https://www.radioid.net/api/dmr/repeater/?country=united%20states', 0)
|
||||
print(result)
|
||||
|
||||
# Try updating subscriber aliases file
|
||||
result = try_download('/tmp/', 'subscribers.json', 'https://www.radioid.net/api/dmr/user/?country=united%20states', 0)
|
||||
print(result)
|
||||
|
||||
# Make Dictionaries
|
||||
peer_ids = mk_id_dict('/tmp/', 'peers.json')
|
||||
if peer_ids:
|
||||
print('ID ALIAS MAPPER: peer_ids dictionary is available')
|
||||
|
||||
subscriber_ids = mk_id_dict('/tmp/', 'subscribers.json')
|
||||
if subscriber_ids:
|
||||
print('ID ALIAS MAPPER: subscriber_ids dictionary is available')
|
||||
|
||||
full_peer_ids = mk_full_id_dict('/tmp/', 'peers.json', 'peer')
|
||||
if peer_ids:
|
||||
print('ID ALIAS MAPPER: full_peer_ids dictionary is available')
|
||||
|
||||
full_subscriber_ids = mk_full_id_dict('/tmp/', 'subscribers.json', 'subscriber')
|
||||
if subscriber_ids:
|
||||
print('ID ALIAS MAPPER: full_subscriber_ids dictionary is available')
|
||||
|
||||
print(get_alias('\x2f\x9b\xe5', subscriber_ids))
|
||||
print(get_info(3120101, subscriber_ids))
|
||||
print(get_alias(3120101, full_subscriber_ids))
|
||||
print(get_info(3120101, full_subscriber_ids))
|
||||
print(get_alias(31201010, subscriber_ids))
|
||||
print(get_info(31201010, subscriber_ids))
|
||||
|
||||
print(get_alias(312000, peer_ids))
|
||||
print(get_info(312000, peer_ids))
|
||||
print(get_alias(312000, full_peer_ids))
|
||||
print(get_info(312000, full_peer_ids))
|
||||
|
||||
print(repr(hex_str_2(65535)))
|
||||
print(repr(hex_str_3(65535)))
|
||||
print(ahex(hex_str_4(13120101)))
|
||||
print(int_id('\x00\xc8\x32\x65'))
|
||||
|
||||
|
||||
|
2
setup.py
2
setup.py
@ -7,7 +7,7 @@ def readme():
|
||||
return file.read()
|
||||
|
||||
setup(name='dmr_utils',
|
||||
version='0.1.19',
|
||||
version='0.1.21',
|
||||
description='ETSI DMR (Digital Mobile Radio) Tier II Utilities',
|
||||
long_description='Modules to disassemble and assemble DMR packets, including generating and decoding various FEC routines',
|
||||
classifiers=[
|
||||
|
Loading…
Reference in New Issue
Block a user