從阿里雲DATAV GeoAtlas接口抽取行政區劃數據

阿里雲提供的地理信息接口

https://datav.aliyun.com/tools/atlas/

有兩個接口, 一個是[行政編碼].json, 一個是[行政編碼]_full.json, 從接口中能夠提取到區縣一級的行政區劃信息. 提取的過程當中遇到的一些問題:

• 從[行政編碼].json中讀取的信息中, 可能parent = null, 出現這種狀況的大都是一些撤縣改區的節點, 要將其設爲上一級節點的行政編碼
• 從[行政編碼].json中讀到的parent的adcode, 可能與[父節點行政編碼]_full.json中讀到的parent的adcode不一致, 例如從110000_full.json中獲得的節點列表, 其parent都是110000, 可是在取其字節點110101.json時會發現, parent變成了110100, 這時候要使用110100這個行政編碼
• 由於從上至下遍歷時, 是不會遇到110100這個節點的, 因此在遍歷的過程當中, 要檢查是否出現了未知的行政編碼, 若是有, 須要額外讀取併入庫
• 有部分節點, 其json沒法讀取(不存在), 例如密雲110118.json, 延慶110119.json, 這時候要用前一步獲得的信息入庫

使用生成的行政區劃數據時, 對於香港澳門的數據, 由於沒有level=city的這一級, 因此須要特殊處理一下, 例如在讀取province這一級的子節點時, 若是發現沒有level=city的節點, 那麼就返回一個虛擬的節點, 這個節點各字段值和本身同樣, 可是level=city.

#!/usr/bin/python3
# -*- coding: UTF-8 -*-

import json
import traceback
import rbcommon

def readRegion(adcode, parent_code = None):
    # https://geo.datav.aliyun.com/areas/bound/140000.json
    url = 'https://geo.datav.aliyun.com/areas/bound/' + adcode + '.json'
    print(url)
    echo = rbcommon.requestGet(url, 'UTF-8', 20, 10)
    if echo is None:
        print('URL request failed: ' + url)
        return
    elif echo.find('<?') == 0:
        print('Not found: ' + url)
        return
    # print(echo)
    json_obj = json.loads(echo)
    region = {}
    region['name'] = json_obj['features'][0]['properties']['name']
    region['adcode'] = json_obj['features'][0]['properties']['adcode']
    region['telecode'] = json_obj['features'][0]['properties']['telecode']
    level = json_obj['features'][0]['properties']['level']
    if (level == 'country'):
        region['level'] = 0
    elif (level == 'province'):
        region['level'] = 1
    elif (level == 'city'):
        region['level'] = 2
    elif (level == 'district'):
        region['level'] = 3
    if ('parent' in json_obj['features'][0]['properties']) and (not json_obj['features'][0]['properties']['parent'] is None):
        region['parent'] = json_obj['features'][0]['properties']['parent']['adcode']
    else:
        region['parent'] = parent_code

    # read sub regions
    sub_regions = []
    region['children'] = sub_regions
    # https://geo.datav.aliyun.com/areas/bound/140000_full.json
    url = 'https://geo.datav.aliyun.com/areas/bound/' + adcode + '_full.json'
    print(url)
    echo = rbcommon.requestGet(url, 'UTF-8', 20, 10)
    if echo is None:
        print('URL request failed: ' + url)
        return region
    elif echo.find('<?') == 0:
        print('Not found: ' + url)
        return region
    # print(echo)
    json_obj = json.loads(echo)
    sub_objs = json_obj['features']
    for sub_obj in sub_objs:
        sub_region = {}
        sub_region['adcode'] = (str)(sub_obj['properties']['adcode'])
        if (sub_region['adcode'] == region['adcode']):
            continue
        sub_region['name'] = sub_obj['properties']['name']
        sub_region['telecode'] = None
        level = sub_obj['properties']['level']
        if (level == 'country'):
            sub_region['level'] = 0
        elif (level == 'province'):
            sub_region['level'] = 1
        elif (level == 'city'):
            sub_region['level'] = 2
        elif (level == 'district'):
            sub_region['level'] = 3
        sub_region['parent'] = adcode
        sub_regions.append(sub_region)

    # further check if the parent adcode is correct
    if (len(sub_regions) > 0):
        # https://geo.datav.aliyun.com/areas/bound/140000.json
        url = 'https://geo.datav.aliyun.com/areas/bound/' + sub_regions[0]['adcode'] + '.json'
        # print(url)
        echo = rbcommon.requestGet(url, 'UTF-8', 20, 10)
        if echo is None:
            print('URL request failed: ' + url)
        elif echo.find('<?') == 0:
            print('Not found: ' + url)
        else:
            json_obj = json.loads(echo)
            if ('parent' in json_obj['features'][0]['properties']) and (not json_obj['features'][0]['properties']['parent'] is None):
                dummy_parent = json_obj['features'][0]['properties']['parent']['adcode']
                if (dummy_parent != sub_regions[0]['parent']):
                    print('Update parent from {} to {}', sub_regions[0]['parent'], dummy_parent)
                    for sub_region in sub_regions:
                        sub_region['parent'] = dummy_parent

    return region

def readAllRegion(parent_region):
    region = readRegion(parent_region['adcode'], parent_region['parent'])
    if not region is None:
        if (not region['parent'] is None) and (not region['parent'] in regions):
            new_region = readRegion(region['parent'], parent_region['parent'])
            if not new_region is None:
                regions.add(new_region['adcode'])
                insert(new_region)

        regions.add(region['adcode'])
        insert(region)

        for sub_region in region['children']:
            readAllRegion(sub_region)
    else:
        regions.add(parent_region['adcode'])
        insert(parent_region)

def insert(region):
    try:
        with rbcommon.mysqlclient.cursor() as cursor:
            sql = 'INSERT IGNORE INTO `s_region` (`id`, `parent_id`, `level`, `name`, `tele_code`, `short_name`, ' \
                  '`full_name`) VALUES (%s, %s, %s, %s, %s, %s, %s)'
            cursor.execute(sql, (
                region['adcode'],
                None if (not 'parent' in region) else region['parent'],
                region['level'],
                region['name'],
                region['telecode'],
                region['name'],
                '{}'))
            rbcommon.mysqlclient.commit()
    except Exception as e:
        print(json.dumps(region))
        traceback.print_exc()

### MAIN ###
regions = set()
region = readRegion('100000')
readAllRegion(region)

其中rbcommon.mysqlclient的初始化方法

mysqlclient = pymysql.connect(
    host=cfg['mysql']['host'],
    port=cfg['mysql']['port'],
    user=cfg['mysql']['user'],
    password=cfg['mysql']['password'],
    db=cfg['mysql']['db'],
    charset=cfg['mysql']['charset'],
    cursorclass=pymysql.cursors.DictCursor)