misp-galaxy/tools/UN M49/generate_region_cluster.py

111 lines
4 KiB
Python
Raw Permalink Normal View History

#!/usr/bin/env python
# -*- coding: utf-8 -*-
import argparse
import csv
import json
from collections import defaultdict
from pathlib import Path
from uuid import uuid4
current_path = Path(__file__).resolve().parent
def _generate_cluster_values(cluster_values: dict):
values = []
for _, cluster_value in sorted(cluster_values.items()):
values.append(cluster_value)
return values
def _parse_csv_input(input_file: Path) -> dict:
# We create a mapping to associate the regions and their sub-regions
regions_mapping = defaultdict(set)
with open(input_file, 'rt', encoding='utf-8') as csvfile:
csv_reader = csv.reader(csvfile, delimiter=',')
next(csv_reader)
for row in csv_reader:
'''
gc: Global Code
gn: Global Name
rc: Region Code
rn: Region Name
src: Sub-region Code
srn: Sub-region Name
irc: Intermediate Region Code
irn: Intermediate Region Name
mc: M49 Code
coa: Country of Area
'''
gc, gn, rc, rn, src, srn, irc, irn, mc, coa, *_ = row
global_region = f'{gc} - {gn}'
if rc and rn:
# Almost all the areas have a region information
region = f'{rc} - {rn}'
regions_mapping[global_region].add(region)
# Deal with the region information
sub_region = f'{src} - {srn}'
regions_mapping[region].add(sub_region)
country = f'{mc} - {coa}'
if irc and irn:
# If the country is located in an intermediate region
inter_region = f'{irc} - {irn}'
# Deal with the sub-region information
regions_mapping[sub_region].add(inter_region)
# Deal with the intermediate region information
regions_mapping[inter_region].add(country)
else:
# The country is located in a sub-region
regions_mapping[sub_region].add(country)
else:
# Should be only Antarctica which has only global region and
# country information
country = f'{mc} - {coa}'
regions_mapping[global_region].add(country)
regions_mapping[country] = None
return regions_mapping
def update_cluster(input_file, filename):
with open(filename, 'rt', encoding='utf-8') as f:
region_cluster = json.load(f)
cluster = {value['value']: value for value in region_cluster['values']}
regions_mapping = _parse_csv_input(input_file)
is_changed = False
for region, subregions in regions_mapping.items():
if region not in cluster:
cluster_value = {
'value': region,
'uuid': uuid4().__str__(),
}
if subregions is not None:
cluster_value['meta'] = {
'subregions': sorted(subregions)
}
cluster[region] = cluster_value
is_changed = True
else:
if subregions != cluster[region].get('meta', {}).get('subregions'):
if subregions is not None:
cluster[region]['meta'] = {
'subregions': sorted(subregions)
}
is_changed = True
if is_changed:
region_cluster['values'] = _generate_cluster_values(cluster)
with open(filename, 'wt', encoding='utf-8') as f:
f.write(json.dumps(region_cluster, indent=2))
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Generates/updates the region galaxy cluster')
parser.add_argument(
'-i', '--input', type=Path, default=current_path / 'UNSD.csv',
help="CSV input file"
)
args = parser.parse_args()
filename = current_path.parents[1] / 'clusters' / 'region.json'
update_cluster(args.input, filename)