guke
dc63632e77
- app/utils/geo.py: reverse_geocoder 单例(mode=1 单进程 KDTree),经纬度→最近聚居点 - app/utils/meituan_city.py: 坐标→美团 city_id(省份/城市名桥接 + 多级兜底 + lru_cache 量化) - feed(rec) / top-sales: 按解析出的 city_id 过滤离线库;城市解析不出 / 老客户端不带坐标 → degraded - top-sales 与 rec 一致置空库内距离(相对城市默认点,对用户无意义) - main.py 启动预热 KDTree;pyproject 加 reverse_geocoder 依赖 + 分发 city_dict.txt - 新增 geo / meituan_city 测试(56 例);scripts/load_meituan_coupon_tsv.py 灌样本到本地 SQLite - .gitignore 忽略样本 TSV 与 .claude 本地设置 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
138 lines
6.0 KiB
Python
138 lines
6.0 KiB
Python
"""reverse_geocoder 经纬度→城市 测试。
|
|
|
|
验证离线库对国内主要城市的匹配准确性。注意：gazetteer 的中国数据粒度不一致——
|
|
直辖市/省会通常直接命中城市名，部分城市可能命中到区/街道级（如天津→Erwangzhuang、
|
|
西安→Zhangjiabao），此时 admin1 为省级行政区。测试以 admin1（省级）匹配为主。
|
|
"""
|
|
from __future__ import annotations
|
|
|
|
import pytest
|
|
|
|
from app.utils.geo import get_city
|
|
|
|
# ─────────────── 国内主要城市 ───────────────
|
|
# (城市, 纬度, 经度, 期望 admin1 包含字串)
|
|
_CITY_CASES = [
|
|
# 直辖市 — admin1 即城市名(可能带 Shi 后缀)
|
|
("北京", 39.9042, 116.4074, "Beijing"),
|
|
("上海", 31.2304, 121.4737, "Shanghai"),
|
|
("重庆", 29.4316, 106.9123, "Chongqing"),
|
|
("天津", 39.3434, 117.3616, "Tianjin"),
|
|
# 省会 / 一线 — admin1 为省份
|
|
("广州", 23.1291, 113.2644, "Guangdong"),
|
|
("深圳", 22.5431, 114.0579, "Guangdong"),
|
|
("成都", 30.5728, 104.0668, "Sichuan"),
|
|
("杭州", 30.2741, 120.1551, "Zhejiang"),
|
|
("武汉", 30.5928, 114.3055, "Hubei"),
|
|
("西安", 34.3416, 108.9398, "Shaanxi"),
|
|
("南京", 32.0603, 118.7969, "Jiangsu"),
|
|
("长沙", 28.2282, 112.9388, "Hunan"),
|
|
("郑州", 34.7466, 113.6253, "Henan"),
|
|
("济南", 36.6512, 116.9946, "Shandong"),
|
|
("昆明", 25.0389, 102.7183, "Yunnan"),
|
|
("福州", 26.0745, 119.2965, "Fujian"),
|
|
("哈尔滨", 45.8038, 126.5350, "Heilongjiang"),
|
|
("乌鲁木齐", 43.8256, 87.6168, "Xinjiang"),
|
|
("拉萨", 29.6500, 91.1000, "Tibet"),
|
|
# 非省会
|
|
("厦门", 24.4798, 118.0894, "Fujian"),
|
|
("苏州", 31.2990, 120.5853, "Jiangsu"),
|
|
("青岛", 36.0671, 120.3826, "Shandong"),
|
|
]
|
|
|
|
|
|
@pytest.mark.parametrize("label,lat,lon,expected_admin1", _CITY_CASES)
|
|
def test_city_admin1_match(label: str, lat: float, lon: float, expected_admin1: str) -> None:
|
|
"""所有城市经纬度应能匹配到正确的省级行政区 (admin1)。"""
|
|
r = get_city(lat, lon)
|
|
assert r["name"] != "", f"{label}: name should not be empty"
|
|
assert r["country"] == "CN", f"{label}: expected country=CN, got={r['country']}"
|
|
assert expected_admin1 in r["admin1"], \
|
|
f"{label}: expected admin1 to contain '{expected_admin1}', got={r['admin1']!r}"
|
|
|
|
|
|
# ─────────────── 直辖市 / 省会直接命中城市名 ───────────────
|
|
# 这些城市在 gazetteer 中的坐标恰好命中城市级条目（而非区/街道级），
|
|
# 验证 name 字段也正确。
|
|
_DIRECT_HIT_CASES = [
|
|
("北京", 39.9042, 116.4074, "Beijing"),
|
|
("上海", 31.2304, 121.4737, "Shanghai"),
|
|
("广州", 23.1291, 113.2644, "Guangzhou"),
|
|
("深圳", 22.5431, 114.0579, "Shenzhen"),
|
|
("成都", 30.5728, 104.0668, "Chengdu"),
|
|
("杭州", 30.2741, 120.1551, "Hangzhou"),
|
|
("郑州", 34.7466, 113.6253, "Zhengzhou"),
|
|
("济南", 36.6512, 116.9946, "Jinan"),
|
|
("昆明", 25.0389, 102.7183, "Kunming"),
|
|
("哈尔滨", 45.8038, 126.5350, "Harbin"),
|
|
("厦门", 24.4798, 118.0894, "Xiamen"),
|
|
("苏州", 31.2990, 120.5853, "Suzhou"),
|
|
("青岛", 36.0671, 120.3826, "Qingdao"),
|
|
("拉萨", 29.6500, 91.1000, "Lhasa"),
|
|
]
|
|
|
|
|
|
@pytest.mark.parametrize("label,lat,lon,expected_name", _DIRECT_HIT_CASES)
|
|
def test_city_name_direct_hit(label: str, lat: float, lon: float, expected_name: str) -> None:
|
|
"""直辖市/省会等主要城市坐标应直接命中城市名(而非区/街道级)。"""
|
|
r = get_city(lat, lon)
|
|
assert r["name"] == expected_name, \
|
|
f"{label}: expected name={expected_name}, got={r['name']!r}"
|
|
|
|
|
|
# ─────────────── 边界情况 ───────────────
|
|
|
|
def test_ocean_not_china() -> None:
|
|
"""远洋坐标不应误判为国内城市。"""
|
|
# 太平洋中部 → 可能匹配到最近有人岛(如法属波利尼西亚 Taiohae)，但绝不应是 CN
|
|
r = get_city(0.0, -140.0)
|
|
assert r["country"] != "CN", f"mid-Pacific should not be CN, got {r}"
|
|
|
|
# 南大西洋
|
|
r2 = get_city(-30.0, -20.0)
|
|
assert r2["country"] != "CN", f"South Atlantic should not be CN, got {r2}"
|
|
|
|
|
|
def test_return_keys_and_types() -> None:
|
|
"""返回 dict 应包含全部五个字段且类型为 str。"""
|
|
r = get_city(39.9042, 116.4074)
|
|
for key in ("name", "admin1", "country", "latitude", "longitude"):
|
|
assert key in r, f"missing key: {key}"
|
|
assert isinstance(r[key], str), f"key {key} should be str, got {type(r[key])}"
|
|
|
|
|
|
def test_empty_result_keys() -> None:
|
|
"""结果始终应包含完整字段且全为 str 类型（即使匹配到偏远地）。"""
|
|
# reverse_geocoder KDTree 总找最近聚居点；业务侧如需判定"是否有效城市"
|
|
# 应自行按 country / admin1 做二次校验，而非依赖空字符串。
|
|
r = get_city(0.0, -140.0)
|
|
assert r["name"] != ""
|
|
assert isinstance(r["name"], str)
|
|
assert isinstance(r["admin1"], str)
|
|
assert isinstance(r["country"], str)
|
|
assert isinstance(r["latitude"], str)
|
|
assert isinstance(r["longitude"], str)
|
|
|
|
|
|
def test_same_coords_consistent() -> None:
|
|
"""同一坐标两次查询应返回相同结果（幂等）。"""
|
|
r1 = get_city(31.2304, 121.4737)
|
|
r2 = get_city(31.2304, 121.4737)
|
|
assert r1 == r2
|
|
|
|
|
|
def test_near_border_has_result() -> None:
|
|
"""省界附近的坐标应返回结果(非空 + 国内)。"""
|
|
# 苏鲁豫皖交界区域(徐州/商丘/宿州附近)
|
|
r = get_city(34.2, 116.8)
|
|
assert r["name"] != "", "border region should find a nearby populated place"
|
|
assert r["country"] == "CN"
|
|
|
|
|
|
def test_extreme_lat_lon_no_crash() -> None:
|
|
"""极值经纬度不应抛异常。"""
|
|
r1 = get_city(90.0, 0.0) # 北极
|
|
r2 = get_city(-90.0, 0.0) # 南极
|
|
assert isinstance(r1, dict)
|
|
assert isinstance(r2, dict)
|