TableOfContents

关于获取要素：

获取要素最好用GetNextFeature而不要用索引值

因为shapefile在实现的时候是以0为起始索引。mapinfo的tab是以1为起始值，这样在获取的时候需要判断读取数据源类型。但是用GetNextFeature就会自动从正确的起始值开始读取。如果需要重新开始读取，就用ResetReading重新设置读取位置到开头。

获取要素中的坐标值最好用ExportToWkb而不要用GetX,GetY

ExportToWkb可以批量获取所有的坐标值。而GetX，GetY只能获取单独的一个值。在绘制的时候还要费神去把它们合成一个数组，而获取了wkb，就可以进行批量解析，一下生成数组，这样，我们就可以利用到了强大的Numeric模块，进行神奇的矩阵魔术。至于wkb的格式，[http://publib.boulder.ibm.com/infocenter/db2luw/v8/index.jsp?topic=/com.ibm.db2.udb.doc/opt/rsbp4121.htm 这里]有简单的说明。 我写了一个小模块，用于解析wkb

import struct
import ogr
from array import array

import sys
endian_name = sys.byteorder

wkbXDR = '>'     # Big Endian
wkbNDR = '<'     # Little Endian

def choose(bool, a, b):
    return (bool and [a] or [b])[0]

BTOR = choose(endian_name == 'little',wkbNDR,wkbNDR)

def up_endian_type(wkb):
    endian_t = struct.unpack('b',wkb[0])[0]
    endian = choose(endian_t,'<','>')
    wkbtype = struct.unpack(endian+'I',wkb[1:5])[0]
    return endian,wkbtype,endian_t

def up_len(wkb,beg,endian):
    return struct.unpack(endian+'I',wkb[beg:beg+4])[0]

def up_point(wkb):
    endian,wkbtype,et = up_endian_type(wkb)
    points = struct.unpack(endian+"2d",wkb[5:])
    return points

def up_linestring(wkb):
    endian,wkbtype,et = up_endian_type(wkb)
    lenght = up_len(wkb,5,endian)
    points = array('d',wkb[9:9+lenght*16])
    if endian != BTOR : points.byteswap()
    return points

def up_linearring(wkb,ringcount,endian):
    #endian,wkbtype,et = up_endian_type(wkb)
    points = []
    ptr = 0
    for i in range(ringcount):
        length = up_len(wkb,ptr,endian)
        ps = array('d',wkb[ptr+4:ptr+4+length*16])
        if endian != BTOR : ps.byteswap()
        points.append(ps)
        ptr += 4+length*16
    return points,ptr

def up_polygon(wkb,sub=-1):
    endian,wkbtype,et = up_endian_type(wkb)
    if sub == -1:
        ringcount = up_len(wkb,5,endian)
        points = up_linearring(wkb[9:],ringcount,endian)[0]
        return points
    else:
        points = []
        ptr = 5
        ringcount = up_len(wkb,ptr,endian)
        ps,ringlen = up_linearring(wkb[ptr+4:],ringcount,endian)
        points.append(ps)
        ptr += 4+ringlen
        return points,ptr

def up_mpoint(wkb):
    endian,wkbtype,et = up_endian_type(wkb)
    subcount = up_len(wkb,5,endian)
    points = []
    ptr = 9
    for i in range(subcount):
        subps = up_point(wkb[ptr:])
        points.append(subps)
        ptr += 9+len(subps)*8
    return points

def up_mlinestring(wkb):
    endian,wkbtype,et = up_endian_type(wkb)
    subcount = up_len(wkb,5,endian)
    points = []
    ptr = 9
    for i in range(subcount):
        subps = up_linestring(wkb[ptr:])
        points.append(subps)
        ptr += 9+len(subps)*8
    return points 

def up_mpolygon(wkb):
    endian,wkbtype,et = up_endian_type(wkb)
    subcount = up_len(wkb,5,endian)
    points = []
    ptr = 9
    for i in range(subcount):
        subps,size = up_polygon(wkb[ptr:],i)
        points.append(subps)
        ptr += size
    return points 

fun_map = {
        ogr.wkbPoint : up_point,
        ogr.wkbLineString : up_linestring,
        ogr.wkbPolygon : up_polygon,
        ogr.wkbMultiPoint : up_mpoint,
        ogr.wkbMultiLineString : up_mlinestring,
        ogr.wkbMultiPolygon : up_mpolygon
        }

def WkbUnPacker(wkb):
    endian,wkbtype,endian_t = up_endian_type(wkb)
    foo = fun_map[wkbtype]
    points = foo(wkb)
    return [endian_t,wkbtype,points]


if __name__ == "__main__":
    import time
    ds = ogr.Open("e:/gisdata/data/country.shp")
    layer = ds.GetLayer()
    begt = time.time()
    #count = layer.GetFeatureCount()
    #count  = 2
    feature = layer.GetNextFeature()
    #for i in [1]:#range(count):
    while feature is not None:
        #feature = layer.GetFeature(i)
        geom = feature.GetGeometryRef()
        #print geom.ExportToWkt()
        wkb = geom.ExportToWkb()
        wkbarr = WkbUnPacker(wkb)
        feature = layer.GetNextFeature()
        #print wkbarr
    print time.time()-begt

因为手上数据不全，无法测试多点、多线、多多边形的情况。如果谁有数据，可以测试一下，如果有错，希望能反馈给我！

加快读取大量feature时的循环速度

ogr在python绑定中读取大量feature时速度总是很慢。16000条记录单单只是获取Feature竟然运行了130秒，简直令人无法忍受。后来打开了ogr.py察看了一下绑定代码，发现在Feature类中有“setarr"和"getarr"两个自省方法在创建Feature后不断运行。试着注释掉它们，速度一下就提高到2秒。简直是飞跃啊！不过注释掉自省方法会不会有后遗症，我现在还不敢说（如果大家发现注释掉这两个自省方法有问题，或者不需要注释掉调代码而有更好的方法，希望能反馈给我）。

反馈

如果您发现我写的东西中有问题，或者您对我写的东西有意见，请一定要发邮件跟我讲，Email( MailTo(linux_23 AT 163 DOT com) )