R：如何将osmdata的运行时减少为igraph转换_R_Openstreetmap_Igraph_Sf_Overpass Api

R：如何将osmdata的运行时减少为igraph转换

R：如何将osmdata的运行时减少为igraph转换,r,openstreetmap,igraph,sf,overpass-api,R,Openstreetmap,Igraph,Sf,Overpass Api,是否可以减少以下代码的运行时间我的目标是从由框边界指定的开放街道数据区域中获取加权igraph对象目前我正在尝试使用OverpassAPI来卸载内存负载，这样我就不必在内存中保存大的osm文件首先，我得到一个由bbox（仅街道）指定为xml结构的osm数据 library(osmdata) library(osmar) install.packages("remotes") remotes::install_github("hypertidy/scgraph&

是否可以减少以下代码的运行时间

我的目标是从由框边界指定的开放街道数据区域中获取加权igraph对象

目前我正在尝试使用OverpassAPI来卸载内存负载，这样我就不必在内存中保存大的osm文件

首先，我得到一个由bbox（仅街道）指定为xml结构的osm数据

library(osmdata)
library(osmar)
install.packages("remotes")
remotes::install_github("hypertidy/scgraph")
library(scgraph)


dat <- opq(bbox = c(11.68771, 47.75233, 12.35058, 48.19743 )) %>%
  add_osm_feature(key = 'highway',value = c("trunk", "trunk_link", "primary","primary_link", "secondary", "secondary_link", "tertiary","tertiary_link", "residential", "unclassified" ))%>% 
  osmdata_xml ()

到一个igraph

速度相当快，但遗憾的是重量正在丢失，所以这不是一个解决方案

原因是：如果我使用函数

osmar:：as_igraph（）

将对象从osmar转换为igraph对象，则权重将根据两条边之间的距离计算并添加到igraph中：

    edges <- lapply(dat, function(x) {
    n <- nrow(x)
    from <- 1:(n - 1)
    to <- 2:n
    weights <- distHaversine(x[from, c("lon", "lat")], x[to, 
      c("lon", "lat")])
    cbind(from_node_id = x[from, "ref"], to_node_id = x[to, 
      "ref"], way_id = x[1, "id"], weights = weights)
  })

edges-sf->igraph

速度更快，但使用这种方法时，我需要根据边的距离将权重合并到图形中，我目前还不能这样做，非常感谢您的帮助

此外，在使用sf和生成的igraph对象时，openstreetmap gps点与其ID之间的连接不应丢失。这意味着我应该能够从生成的Igraph中找到ID的gps位置。一个查找表就足够了。如果我走

overpass->silicate->igraph

或

overpass->xml->osmar->igraph

路线，这是可能的。我不确定是否仍然可以使用

overpass->sf->igraph

route。将xml数据转换为另一种格式似乎需要很长时间。与其使用xml，不如让overpass返回一个

sf

对象，并使用它可能会更快。然后，

sf

对象可以被

sfnetworks

包操作和使用以获得网络，同时保留网络的空间方面。可以通过来自

sfnetworks

（或

tidygraph

）包的函数添加权重

我认为下面的代码重点关注速度和边缘权重问题。其他问题，如查找最近的节点或边，可以使用

sf

包的函数解决，但没有解决。否则，这将不仅仅是一个一次性的SO问题

通过对边缘使用

st_simplify

，可以以空间精度为代价提高速度。这种方法的一个问题是stnetworks将一个节点放置在每个线串与另一个线串相交的位置。返回的数据通常将一条道路分割为多个线串。作为示例，请参见下图边缘上两条较长的黄色道路。可能是一个可以解决的问题，但在这种情况下可能不值得

library(osmdata)
#library(osmar)
library(tidyverse)
library(sf)
library(ggplot2)
library(sfnetworks)
library(tidygraph)


# get data as an sf object rather than xml
## This is the slowest part of the code.
dat_sf <- opq(bbox = c(11.68771, 47.75233, 12.35058, 48.19743 )) %>%
  add_osm_feature(key = 'highway',value = c("trunk", "trunk_link", "primary","primary_link", "secondary", "secondary_link", "tertiary","tertiary_link", "residential", "unclassified" ))%>% 
  osmdata_sf()



# Only keep lines & polygons. Points take up too much memory & 
## all seem to be on lines anyway. Change polygons to LINESTRING,
## as most seem to be roundabouts or a few odd cases.
lines_sf <- dat_sf$osm_lines %>% select(osm_id) %>% st_sf()
polys_sf <- dat_sf$osm_polygons %>% select(osm_id) %>% st_sf() %>% 
  st_cast('LINESTRING')

# Combine the two above sf objects into one 
dat_sf_bound <- rbind(lines_sf, polys_sf)

# get an sfnetwork
dat_sf_net <- as_sfnetwork(dat_sf_bound)

# add edge weight as distance
dat_sf_net <- dat_sf_net %>%
  activate(edges) %>%
  mutate(weight = edge_length())

仅边和带节点边的打印：

有几条长路会使颜色倾斜，但说明了将一条路一分为二的效果

编辑：寻址注释以按纬度/经度坐标选择最近的边（道路）。节点（上面的交点/红点）没有我知道的osm id号。节点由

sfnetworks

创建

从lat/lon点的

sf

对象开始，作为我们制作的gps坐标

# random point
gps <- sfheaders::sf_point(data.frame(x = 11.81854, y = 48.04514)) %>% st_set_crs(4326)

# nearest edge(road) to the point. dat_sf_net must have edges activated.
near_edge <- st_nearest_feature(gps, dat_sf_net %>% st_as_sf())

>near_edge
[1] 4359

> st_as_sf(dat_sf_net)[near_edge,]
Simple feature collection with 1 feature and 4 fields
Geometry type: LINESTRING
Dimension:     XY
Bounding box:  xmin: 11.81119 ymin: 48.02841 xmax: 11.82061 ymax: 48.06845
Geodetic CRS:  WGS 84
# A tibble: 1 x 5
   from    to osm_id                                                           geometry   weight
  <int> <int> <chr>                                                    <LINESTRING [°]>      [m]
1  7590  7591 24232418 (11.81289 48.02841, 11.81213 48.03014, 11.81202 48.03062, 11.81… 4576.273

p3 <- gplot() +
      geom_sf(data = st_as_sf(dat_sf_net), color = 'black') +       
      geom_sf(data = gps, color = 'red') + 
      geom_sf(data = st_as_sf(dat_sf_net)[near_edge,], color = 'orange') + 
      coord_sf(xlim = c(11.7, 11.9), ylim = c(48, 48.1))

#随机点
全球定位系统%st\U set\U crs（4326）
#距离该点最近的边缘（道路）。dat_sf_网络必须激活边缘。
靠近边缘%st\u作为\u sf（））
>近边缘
[1] 4359
>st_as_sf（dat_sf_net）[近边缘，]
具有1个要素和4个字段的简单要素集合
几何图形类型：线条字符串
尺寸：XY
边界框：xmin:11.81119 ymin:48.02841 xmax:11.82061 ymax:48.06845
大地测量CRS:WGS 84
#一个tibble:1 x 5
从到osm_id几何体权重
[m]
1  7590  7591 24232418 (11.81289 48.02841, 11.81213 48.03014, 11.81202 48.03062, 11.81… 4576.273
p3如果要从R中的道路网络开始创建图形对象，则我将使用以下步骤
首先，我需要从githubrepo安装sfnetworks
（因为我们最近修复了一些bug，而且最新版本不在CRAN上）
remotes:：install_github（“luukvdmeer/sfnetworks”，quiet=TRUE）

然后加载包
库（sf）
#>链接到GEOS 3.9.0、GDAL 3.2.1、项目7.2.1
图书馆（潮汐图）
#> 
#>附加包：“tidygraph”
#>以下对象已从“package:stats”屏蔽：
#> 
#>滤器
图书馆（SF网络）
图书馆（osmdata）
#>数据（c）OpenStreetMap贡献者，ODbL 1.0。https://www.openstreetmap.org/copyright

从立交桥API下载数据
my_osm_数据%
添加osm功能(
键=‘高速公路’，
值=c（“中继”、“中继链路”、“主”、“主链路”、“次”、“次链路”、“次链路”、“三级”、“三级链路”、“住宅”、“未分类”）
) %>% 
osmdata_sf（安静=假）
#>正在向立交桥API发出查询。。。
#>利率上限：2
#>查询完成！
#>将OSM数据转换为sf格式

现在，我提取道路并构建sfnetwork对象：
system.time({
#拔除道路
my_roads 3.03 0.16 3.28

如您所见，在下载OSM数据后，运行该过程只需几秒钟
目前，我忽略了my_osm_data$osm_line
中的所有字段，但是如果您需要将my_osm_data$osm_line
中的一些列添加到my_roads
，那么您可以修改前面的代码如下：my roads嗨！如果您对osmar
的替代方法感兴趣，我可以尝试提供一个基于名为.sfnetworks
的R包的解决方案基于tidygraph
，这意味着sfnetworks
返回的对象也是igraph对象。您好，绝对是。我需要能够做两件事：1.基于bbox区域从立交桥获取数据，2.从该区域获取igraph 3.be能耐
    edges <- lapply(dat, function(x) {
    n <- nrow(x)
    from <- 1:(n - 1)
    to <- 2:n
    weights <- distHaversine(x[from, c("lon", "lat")], x[to, 
      c("lon", "lat")])
    cbind(from_node_id = x[from, "ref"], to_node_id = x[to, 
      "ref"], way_id = x[1, "id"], weights = weights)
  })

library(osmdata)
#library(osmar)
library(tidyverse)
library(sf)
library(ggplot2)
library(sfnetworks)
library(tidygraph)


# get data as an sf object rather than xml
## This is the slowest part of the code.
dat_sf <- opq(bbox = c(11.68771, 47.75233, 12.35058, 48.19743 )) %>%
  add_osm_feature(key = 'highway',value = c("trunk", "trunk_link", "primary","primary_link", "secondary", "secondary_link", "tertiary","tertiary_link", "residential", "unclassified" ))%>% 
  osmdata_sf()



# Only keep lines & polygons. Points take up too much memory & 
## all seem to be on lines anyway. Change polygons to LINESTRING,
## as most seem to be roundabouts or a few odd cases.
lines_sf <- dat_sf$osm_lines %>% select(osm_id) %>% st_sf()
polys_sf <- dat_sf$osm_polygons %>% select(osm_id) %>% st_sf() %>% 
  st_cast('LINESTRING')

# Combine the two above sf objects into one 
dat_sf_bound <- rbind(lines_sf, polys_sf)

# get an sfnetwork
dat_sf_net <- as_sfnetwork(dat_sf_bound)

# add edge weight as distance
dat_sf_net <- dat_sf_net %>%
  activate(edges) %>%
  mutate(weight = edge_length())

> dat_sf_net
# An sfnetwork with 33255 nodes and 28608 edges
#
# CRS:  EPSG:4326 
#
# A directed multigraph with 6391 components with spatially explicit edges
#
# Edge Data:     28,608 x 4 (active)
# Geometry type: LINESTRING
# Dimension:     XY
# Bounding box:  xmin: 11.6757 ymin: 47.74745 xmax: 12.39161 ymax: 48.22025
   from    to   weight                                                                  geometry
  <int> <int>      [m]                                                          <LINESTRING [°]>
1     1     2 306.3998 (11.68861 47.90971, 11.6878 47.90965, 11.68653 47.90954, 11.68597 47.909…
2     3     4 245.9225 (11.75216 48.17638, 11.75224 48.17626, 11.75272 48.17556, 11.7528 48.175…
3     5     6 382.2423 (11.7528 48.17351, 11.75264 48.17344, 11.75227 48.17329, 11.752 48.1732,…
4     7     8 131.1373 (11.70029 47.87861, 11.70046 47.87869, 11.70069 47.87879, 11.70138 47.87…
5     9    10 252.9170 (11.75733 48.17339, 11.75732 48.17343, 11.75726 48.17357, 11.75718 48.17…
6    11    12 131.6942 (11.75582 48.17036, 11.75551 48.1707, 11.75521 48.17106, 11.75507 48.171…
# … with 28,602 more rows
#
# Node Data:     33,255 x 1
# Geometry type: POINT
# Dimension:     XY
# Bounding box:  xmin: 11.6757 ymin: 47.74745 xmax: 12.39161 ymax: 48.22025
             geometry
          <POINT [°]>
1 (11.68861 47.90971)
2 (11.68454 47.90937)
3 (11.75216 48.17638)
# … with 33,252 more rows

# random point
gps <- sfheaders::sf_point(data.frame(x = 11.81854, y = 48.04514)) %>% st_set_crs(4326)

# nearest edge(road) to the point. dat_sf_net must have edges activated.
near_edge <- st_nearest_feature(gps, dat_sf_net %>% st_as_sf())

>near_edge
[1] 4359

> st_as_sf(dat_sf_net)[near_edge,]
Simple feature collection with 1 feature and 4 fields
Geometry type: LINESTRING
Dimension:     XY
Bounding box:  xmin: 11.81119 ymin: 48.02841 xmax: 11.82061 ymax: 48.06845
Geodetic CRS:  WGS 84
# A tibble: 1 x 5
   from    to osm_id                                                           geometry   weight
  <int> <int> <chr>                                                    <LINESTRING [°]>      [m]
1  7590  7591 24232418 (11.81289 48.02841, 11.81213 48.03014, 11.81202 48.03062, 11.81… 4576.273

p3 <- gplot() +
      geom_sf(data = st_as_sf(dat_sf_net), color = 'black') +       
      geom_sf(data = gps, color = 'red') + 
      geom_sf(data = st_as_sf(dat_sf_net)[near_edge,], color = 'orange') + 
      coord_sf(xlim = c(11.7, 11.9), ylim = c(48, 48.1))


my_sfn
#> # A sfnetwork with 33179 nodes and 28439 edges
#> #
#> # CRS:  EPSG:4326 
#> #
#> # An undirected multigraph with 6312 components with spatially explicit edges
#> #
#> Registered S3 method overwritten by 'cli':
#>   method     from         
#>   print.boxx spatstat.geom
#> # Node Data:     33,179 x 1 (active)
#> # Geometry type: POINT
#> # Dimension:     XY
#> # Bounding box:  xmin: 11.6757 ymin: 47.74745 xmax: 12.39161 ymax: 48.22025
#>                     x
#>           <POINT [°]>
#> 1 (11.68861 47.90971)
#> 2 (11.68454 47.90937)
#> 3 (11.75216 48.17638)
#> 4 (11.75358 48.17438)
#> 5  (11.7528 48.17351)
#> 6 (11.74822 48.17286)
#> # ... with 33,173 more rows
#> #
#> # Edge Data:     28,439 x 4
#> # Geometry type: LINESTRING
#> # Dimension:     XY
#> # Bounding box:  xmin: 11.6757 ymin: 47.74745 xmax: 12.39161 ymax: 48.22025
#>    from    to                                                           x weight
#>   <int> <int>                                            <LINESTRING [°]>  <dbl>
#> 1     1     2 (11.68861 47.90971, 11.6878 47.90965, 11.68653 47.90954, 1~   306.
#> 2     3     4 (11.75216 48.17638, 11.75224 48.17626, 11.75272 48.17556, ~   246.
#> 3     5     6 (11.7528 48.17351, 11.75264 48.17344, 11.75227 48.17329, 1~   382.
#> # ... with 28,436 more rows

class(my_sfn)
#> [1] "sfnetwork" "tbl_graph" "igraph"

as.igraph(my_sfn)
#> IGRAPH 101dcdf U-W- 33179 28439 -- 
#> + attr: x (v/x), x (e/x), weight (e/n)
#> + edges from 101dcdf:
#>  [1]   1--  2   3--  4   5--  6   7--  8   9-- 10  11-- 12  13-- 14  15-- 16
#>  [9]  17-- 18  16-- 19  20-- 21  21-- 22  23-- 24  25-- 26  27-- 28  29-- 30
#> [17]  31-- 32  33-- 34  35-- 36  37-- 38  39-- 40  41-- 42  43-- 44  45-- 46
#> [25]  14-- 47  48-- 49  50-- 51  52-- 53  54-- 55  56-- 57  36-- 58  58-- 59
#> [33]  60-- 61  62-- 63  64-- 65  66-- 67  68-- 69  70-- 71  72-- 73  74-- 75
#> [41]  76-- 77  78-- 79  80-- 81  82-- 83  84-- 85  86-- 87  88-- 89  90-- 91
#> [49]  92-- 93  94-- 95  96-- 97  98-- 99 100--101 102--103 104--105 106--107
#> [57] 108--109 110--111 112--113  80--114 115--116 117--118 119--120 121--122
#> + ... omitted several edges

all.equal(
  target = igraph::edge_attr(as.igraph(my_sfn), "weight"), 
  current = as.numeric(st_length(my_roads))
)
#> [1] TRUE