使用 mssql-python 处理空间数据

Microsoft SQL 提供了两种空间数据类型，您可以通过 mssql-python 驱动使用。 根据你的坐标代表的类型选择：

插入空间数据

使用 Microsoft SQL 的构造函数（如 geography::Point()），或传递众所周知文本 (WKT) 字符串。

地理数据（点）

使用带有纬度、经度和SRID的点构造器插入地理点。

import mssql_python

connection_string = "Server=<server>.database.windows.net;Database=AdventureWorks2022;Authentication=ActiveDirectoryDefault;Encrypt=yes"

conn = mssql_python.connect(connection_string)
cursor = conn.cursor()

# Insert a geographic point (longitude, latitude)
# Note: Microsoft SQL uses (longitude, latitude) order
cursor.execute("CREATE TABLE #Locations (Name NVARCHAR(100), GeoLocation GEOGRAPHY)")
cursor.execute("""
    INSERT INTO #Locations (Name, GeoLocation)
    VALUES (%(name)s, geography::Point(%(lat)s, %(lon)s, 4326))
""", {"name": "Seattle", "lat": 47.6062, "lon": -122.3321})
conn.commit()

根据 WKT（众所周知文本）创建地理对象

使用 Well-Known Text 格式插入地理数据，该格式支持点、线串和多边形。

# Well-Known Text format
wkt_point = "POINT(-122.3321 47.6062)"
wkt_line = "LINESTRING(-122.3321 47.6062, -122.4194 37.7749)"
wkt_polygon = "POLYGON((-122.40 47.60, -122.30 47.60, -122.30 47.65, -122.40 47.65, -122.40 47.60))"

cursor.execute("DROP TABLE IF EXISTS #Locations")
cursor.execute("CREATE TABLE #Locations (Name NVARCHAR(100), GeoLocation GEOGRAPHY)")
cursor.execute("""
    INSERT INTO #Locations (Name, GeoLocation)
    VALUES (%(name)s, geography::STGeomFromText(%(wkt)s, 4326))
""", {"name": "Route", "wkt": wkt_line})
conn.commit()

几何数据

使用点和多边形插入平面几何数据，用于 CAD 图纸或平面图。

# Insert a geometry point (flat coordinate system)
cursor.execute("CREATE TABLE #FloorPlan (RoomName NVARCHAR(100), RoomShape GEOMETRY)")
cursor.execute("""
    INSERT INTO #FloorPlan (RoomName, RoomShape)
    VALUES (%(name)s, geometry::Point(%(x)s, %(y)s, 0))
""", {"name": "Office 101", "x": 50.0, "y": 100.0})

# Insert a geometry polygon
room_wkt = "POLYGON((0 0, 0 10, 20 10, 20 0, 0 0))"
cursor.execute("""
    INSERT INTO #FloorPlan (RoomName, RoomShape)
    VALUES (%(name)s, geometry::STGeomFromText(%(wkt)s, 0))
""", {"name": "Conference Room", "wkt": room_wkt})
conn.commit()

查询空间数据

使用像 STAsText() 和 属性 Lat/Long 这样的空间方法，以可读格式检索坐标。

以文本形式检索

以 Well-Known Text（WKT）格式检索空间坐标，以及纬度和经度属性。

cursor.execute("""
    SELECT 
        City,
        SpatialLocation.STAsText() AS LocationWKT,
        SpatialLocation.Lat AS Latitude,
        SpatialLocation.Long AS Longitude
    FROM Person.Address
    WHERE SpatialLocation IS NOT NULL
""")

for row in cursor.fetchall()[:5]:
    print(f"{row.City}: {row.Latitude}, {row.Longitude}")
    print(f"  WKT: {row.LocationWKT}")

以 GeoJSON 格式检索

Microsoft SQL 支持 GeoJSON 转换。 在 SQL Server 2017+ 中，您可以使用字符串函数直接构建 GeoJSON，或像下面所示的 Python 转换：

cursor.execute("""
    SELECT 
        City,
        SpatialLocation.STAsText() AS WKT
    FROM Person.Address
    WHERE SpatialLocation IS NOT NULL AND City = %(city)s
""", {"city": "Seattle"})

for row in cursor:
    print(f"{row.City}: {row.WKT}")

查找附近的点

利用空间距离函数查找距离参考点指定距离内的所有位置。

# Find locations within 10 km of Seattle
cursor.execute("""
    DECLARE @seattle geography = geography::Point(47.6062, -122.3321, 4326);
    
    SELECT 
        City,
        SpatialLocation.STDistance(@seattle) / 1000 AS DistanceKM
    FROM Person.Address
    WHERE SpatialLocation IS NOT NULL
      AND SpatialLocation.STDistance(@seattle) < 50000  -- 50 km in meters
    ORDER BY SpatialLocation.STDistance(@seattle)
""")

for row in cursor.fetchall()[:5]:
    print(f"{row.City}: {row.DistanceKM:.2f} km away")

在多边形内寻找点

找到所有与地理多边形区域相交或落在该区域内的点。

# Find all addresses within a region
cursor.execute("""
    DECLARE @region geography = geography::STPolyFromText(
        'POLYGON((-122.5 47.5, -122.2 47.5, -122.2 47.7, -122.5 47.7, -122.5 47.5))',
        4326
    );
    
    SELECT City, SpatialLocation.STAsText() AS Location
    FROM Person.Address
    WHERE SpatialLocation IS NOT NULL
      AND @region.STIntersects(SpatialLocation) = 1
""")

计算距离

Microsoft SQL 的 STDistance() 方法对于 geography 类型返回以米为单位的距离。

两点之间的距离

创建一个辅助函数来计算两个地理点之间的公里距离。

def get_distance_km(cursor, point1: tuple, point2: tuple) -> float:
    """Calculate distance between two points in kilometers."""
    cursor.execute("""
        DECLARE @point1 geography = geography::Point(%(lat1)s, %(lon1)s, 4326);
        DECLARE @point2 geography = geography::Point(%(lat2)s, %(lon2)s, 4326);
        SELECT @point1.STDistance(@point2) / 1000 AS DistanceKM;
    """, {
        "lat1": point1[0], "lon1": point1[1],
        "lat2": point2[0], "lon2": point2[1]
    })
    return cursor.fetchval()

# Seattle to San Francisco
distance = get_distance_km(cursor, (47.6062, -122.3321), (37.7749, -122.4194))
print(f"Distance: {distance:.2f} km")

计算面积

计算一个地理区域的面积（单位为平方公里）。

cursor.execute("""
    DECLARE @region geography = geography::STPolyFromText(
        'POLYGON((-122.5 47.5, -122.2 47.5, -122.2 47.7, -122.5 47.7, -122.5 47.5))',
        4326
    );
    SELECT 
        'Seattle Region' AS Name,
        @region.STArea() / 1000000 AS AreaSqKm
""")

row = cursor.fetchone()
print(f"{row.Name}: {row.AreaSqKm:.2f} sq km")

空间操作

Microsoft SQL 支持对空间对象进行集合操作，包括并集、交集和缓冲区。

形状的并集

将两个地理多边形合并为一个形状，计算总面积。

cursor.execute("""
    DECLARE @parcel1 geography = geography::STPolyFromText(
        'POLYGON((-122.35 47.60, -122.33 47.60, -122.33 47.62, -122.35 47.62, -122.35 47.60))',
        4326
    );
    DECLARE @parcel2 geography = geography::STPolyFromText(
        'POLYGON((-122.34 47.61, -122.32 47.61, -122.32 47.63, -122.34 47.63, -122.34 47.61))',
        4326
    );
    DECLARE @combined geography = @parcel1.STUnion(@parcel2);
    
    SELECT @combined.STAsText() AS CombinedWKT,
           @combined.STArea() / 1000000 AS TotalAreaKM
""")

row = cursor.fetchone()
print(f"Combined area: {row.TotalAreaKM:.2f} sq km")

路口

找到两个地理区域交汇的重叠区域。

polygon1_wkt = "POLYGON((-122.35 47.60, -122.33 47.60, -122.33 47.62, -122.35 47.62, -122.35 47.60))"
polygon2_wkt = "POLYGON((-122.34 47.61, -122.32 47.61, -122.32 47.63, -122.34 47.63, -122.34 47.61))"

cursor.execute("""
    DECLARE @region1 geography = geography::STPolyFromText(%(wkt1)s, 4326);
    DECLARE @region2 geography = geography::STPolyFromText(%(wkt2)s, 4326);
    
    SELECT 
        @region1.STIntersection(@region2).STAsText() AS IntersectionWKT,
        @region1.STIntersection(@region2).STArea() / 1000000 AS AreaKM
""", {"wkt1": polygon1_wkt, "wkt2": polygon2_wkt})

缓冲区（扩展面积）

在一个地理点周围创建一个缓冲区，找到缓冲半径内的所有位置。

# Find all addresses within 5km buffer of a point
cursor.execute("""
    DECLARE @center geography = geography::Point(47.6062, -122.3321, 4326);
    DECLARE @buffer geography = @center.STBuffer(5000);  -- 5 km buffer
    
    SELECT City
    FROM Person.Address
    WHERE SpatialLocation IS NOT NULL
      AND @buffer.STIntersects(SpatialLocation) = 1
""")

Python 集成

从Microsoft SQL检索WKT字符串，并用Shapely库解析它们以实现客户端几何处理。

与Shapely图书馆合作

运行 pip install shapely 进行安装。

from shapely import wkt
from shapely.geometry import Point, Polygon

# Retrieve spatial data from Person.Address
cursor.execute("""
    SELECT City, SpatialLocation.STAsText() AS WKT
    FROM Person.Address
    WHERE SpatialLocation IS NOT NULL AND City IN ('Seattle', 'Redmond')
""")

for row in cursor:
    # Parse WKT into Shapely geometry
    geom = wkt.loads(row.WKT)
    
    if isinstance(geom, Point):
        print(f"{row.City}: Point at ({geom.x}, {geom.y})")
    elif isinstance(geom, Polygon):
        print(f"{row.City}: Polygon with area {geom.area}")

用 Shapely 创建空间数据

使用 Shapely 库创建几何对象，并将其转换为 WKT 格式，插入 Microsoft SQL 中。

from shapely.geometry import Point, Polygon, LineString
from shapely import wkt

# Create geometries in Python
seattle = Point(-122.3321, 47.6062)
seattle_wkt = wkt.dumps(seattle)

route = LineString([(-122.3321, 47.6062), (-122.4194, 37.7749)])
route_wkt = wkt.dumps(route)

# Insert into a temp table
cursor.execute("CREATE TABLE #Routes (Name NVARCHAR(100), Path NVARCHAR(MAX))")
cursor.execute("""
    INSERT INTO #Routes (Name, Path)
    VALUES (%(name)s, %(wkt)s)
""", {"name": "Seattle to SF", "wkt": route_wkt})

cursor.execute("SELECT Name, Path FROM #Routes")
row = cursor.fetchone()
print(f"{row.Name}: {row.Path[:40]}...")

转换为 GeoJSON

将空间数据从 WKT 格式转换为 GeoJSON，适用于基于网页的应用和地图服务。

import json
from shapely import wkt
from shapely.geometry import mapping

cursor.execute("""
    SELECT City, SpatialLocation.STAsText() AS WKT
    FROM Person.Address
    WHERE SpatialLocation IS NOT NULL AND City = 'Seattle'
""")

row = cursor.fetchone()
geom = wkt.loads(row.WKT)

geojson = {
    "type": "Feature",
    "properties": {"name": row.City},
    "geometry": mapping(geom)
}

print(json.dumps(geojson, indent=2))

GeoDataFrame 集成

将Microsoft SQL的空间数据加载到GeoPandas的GeoDataFrame中，进行高级地理空间分析。

import geopandas as gpd
from shapely import wkt
import pandas as pd

cursor.execute("""
    SELECT AddressID, City, SpatialLocation.STAsText() AS WKT
    FROM Person.Address
    WHERE SpatialLocation IS NOT NULL AND City = 'Seattle'
""")

# Build DataFrame
rows = cursor.fetchall()
df = pd.DataFrame(
    [(r.AddressID, r.City, r.WKT) for r in rows],
    columns=['id', 'city', 'wkt']
)

# Convert to GeoDataFrame
df['geometry'] = df['wkt'].apply(wkt.loads)
gdf = gpd.GeoDataFrame(df, geometry='geometry', crs="EPSG:4326")

# Now use GeoPandas operations
print(gdf.head())

空间索引

在 Microsoft SQL 中创建空间索引，以加速对大型空间数据集的查询。

-- Geography index on Person.Address
CREATE SPATIAL INDEX SIX_Address_SpatialLocation
ON Person.Address(SpatialLocation)
USING GEOGRAPHY_GRID
WITH (
    GRIDS = (LEVEL_1 = MEDIUM, LEVEL_2 = MEDIUM, LEVEL_3 = MEDIUM, LEVEL_4 = MEDIUM),
    CELLS_PER_OBJECT = 16
);

-- Geometry index (example with custom table)
CREATE SPATIAL INDEX SIX_FloorPlan_RoomShape
ON dbo.FloorPlan(RoomShape)
USING GEOMETRY_GRID
WITH (
    BOUNDING_BOX = (0, 0, 1000, 1000),
    GRIDS = (LEVEL_1 = HIGH, LEVEL_2 = HIGH, LEVEL_3 = HIGH, LEVEL_4 = HIGH)
);

性能提示

应用这些技术以保持大规模空间查询的高效性。

使用空间索引提示

利用空间索引加速对大型空间数据集的查询。

cursor.execute("""
    DECLARE @region geography = geography::STPolyFromText(
        'POLYGON((-122.5 47.5, -122.2 47.5, -122.2 47.7, -122.5 47.7, -122.5 47.5))',
        4326
    );

    SELECT City
    FROM Person.Address
    WHERE SpatialLocation IS NOT NULL
      AND SpatialLocation.STIntersects(@region) = 1
""")

先滤波，再计算

先应用快速边界框滤波器，然后进行精确距离计算，从而优化空间查询。

# Approximate filter with bounding box, then precise calculation
cursor.execute("""
    DECLARE @center geography = geography::Point(47.6062, -122.3321, 4326);
    
    SELECT City, SpatialLocation.STDistance(@center) AS Distance
    FROM Person.Address
    WHERE SpatialLocation IS NOT NULL
      AND SpatialLocation.Filter(@center.STBuffer(10000)) = 1  -- Fast bounding box filter
      AND SpatialLocation.STDistance(@center) < 10000         -- Precise distance check
    ORDER BY Distance
""")

降低显示精度

通过降低坐标精度来简化显示空间几何。

cursor.execute("""
    SELECT 
        City,
        SpatialLocation.Reduce(100).STAsText() AS SimplifiedWKT  -- 100 meter tolerance
    FROM Person.Address
    WHERE SpatialLocation IS NOT NULL AND City = 'Seattle'
""")

坐标参考系统

SRID 确定坐标系，并影响 Microsoft SQL 计算距离和面积的方式。

常见SRID值

SRID（空间参考标识符）定义了你数据的坐标系。 使用错误的SRID会导致距离和面积计算错误。

在 SRID 之间转换

获取空间数据并检查其SRID以验证坐标系。

cursor.execute("""
    -- Geography is always round-earth, but SRID defines datum
    SELECT SpatialLocation.STAsText() AS WKT,
           SpatialLocation.STSrid AS SRID
    FROM Person.Address
    WHERE SpatialLocation IS NOT NULL
""")

最佳做法

应用这些指导方针，正确高效地处理空间数据。

验证几何形状

在处理前检查空间数据的有效性并识别任何几何问题。

cursor.execute("""
    SELECT 
        City,
        SpatialLocation.STIsValid() AS IsValid,
        SpatialLocation.IsValidDetailed() AS InvalidReason
    FROM Person.Address
    WHERE SpatialLocation IS NOT NULL
      AND SpatialLocation.STIsValid() = 0
""")

for row in cursor:
    print(f"Invalid: {row.City} - {row.InvalidReason}")

使几何体有效

使用该 MakeValid() 方法自动纠正无效几何形状。

# Example with a temp table
cursor.execute("""
    CREATE TABLE #SpatialFix (Name NVARCHAR(100), GeoLocation GEOGRAPHY);
    INSERT INTO #SpatialFix VALUES ('Test', geography::STGeomFromText('POLYGON((0 0, 0 1, 1 0, 0 0))', 4326));
    UPDATE #SpatialFix
    SET GeoLocation = GeoLocation.MakeValid()
    WHERE GeoLocation.STIsValid() = 0
""")

选择合适的类型

与 geography 之间的geometry选择决定了 Microsoft SQL 如何计算距离和面积：

• 地理：现实世界位置（GPS点）、地表计算、以米为单位的距离。
• 几何形状：平面（平面图、CAD）、笛卡尔坐标系，或当SRID不重要时。

相关内容

• 数据类型映射
• Microsoft SQL 空间文档