首页 > 解决方案 > 带有索引列表的 R tibble:如何快速使用它们?

问题描述

我正在寻找一种快速方法来根据另一个表中的索引列表获取表中列的总和。

这是一个可重现的简单示例:首先创建一个边缘表

fake_edges <- st_sf(data.frame(id=c('a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i'),
                               weight=c(102.1,98.3,201.0,152.3,176.4,108.6,151.4,186.3,191.2), 
                               soc=c(-0.1,0.7,1.1,0.2,0.5,-0.2,0.4,0.3,0.8), 
                               geometry=st_sfc(st_linestring(rbind(c(1,1), c(1,2))),
                                               st_linestring(rbind(c(1,2), c(2,2))),
                                               st_linestring(rbind(c(2,2), c(2,3))),
                                               st_linestring(rbind(c(1,1), c(2,1))),
                                               st_linestring(rbind(c(2,1), c(2,2))),
                                               st_linestring(rbind(c(2,2), c(3,2))),
                                               st_linestring(rbind(c(1,1), c(1,0))),
                                               st_linestring(rbind(c(1,0), c(0,0))),
                                               st_linestring(rbind(c(0,0), c(0,1)))
                                              )))

tm_shape(fake_edges, ext = 1.3) +
 tm_lines(lwd = 2) +
tm_shape(st_cast(fake_edges, "POINT")) +
  tm_dots(size = 0.3) +
tm_graticules(lines = FALSE)

在此处输入图像描述

然后从表中创建一个网络,并找到从第一个节点到所有节点的成本最低的路径。

fake_net <- as_sfnetwork(fake_edges)

fake_paths <- st_network_paths(fake_net,
                         from=V(fake_net)[1],
                         to=V(fake_net),
                         weights='weight', type='shortest')

现在,我要改进的是为该fake_paths表的每一行查找的过程

我所做的是以下(这里有 9 行很快,但在大型网络上需要很长时间):

# Transforming to data.tables makes things a bit faster
fake_p <- as.data.table(fake_paths)
fake_e <- as.data.table(fake_edges)
# ID of the last edge on the path
fake_p$id <- apply(fake_p, 1, function(df) unlist(fake_e[df$edge_paths %>% last(), 'id'], use.names=F))
# Sum of soc
fake_p$result <- to_vec(for (edge in 1:nrow(fake_p)) fake_e[unlist(fake_p[edge, 'edge_paths']), soc] %>% sum())

最终,我想要的是soc我要求result加入的总和支持原版fake_edges

fake_e = left_join(fake_e, 
                   fake_p %>% select(id, result) %>% drop_na(id) %>% mutate(id=as.character(id), result=as.numeric(result)),
                   by='id')
fake_edges$result <- fake_e$result
fake_edges

Simple feature collection with 9 features and 4 fields
Geometry type: LINESTRING
Dimension:     XY
Bounding box:  xmin: 0 ymin: 0 xmax: 3 ymax: 3
CRS:           NA
ID 重量 社会 几何学 结果
一种 102.1 -0.1 线串 (1 1, 1 2) -0.1
b 98.3 0.7 线串 (1 2, 2 2) 0.6
C 201.0 1.1 线串 (2 2, 2 3) 1.7
d 152.3 0.2 线串 (1 1, 2 1) 0.2
e 176.4 0.5 线串 (2 1, 2 2) 不适用
F 108.6 -0.2 线串 (2 2, 3 2) 0.4
G 151.4 0.4 线串 (1 1, 1 0) 0.4
H 186.3 0.3 线串 (1 0, 0 0) 0.7
一世 191.2 0.8 线串 (0 0, 0 1) 1.5

标签: rloopsshortest-pathtibblesfnetwork

解决方案


我不确定您要完成什么,但以下过程应该与您在第一篇文章中描述的过程相对应。

加载包

suppressPackageStartupMessages({
  library(sf)
  library(igraph)
  library(tidygraph)
  library(sfnetworks)
  library(tibble)
})

定义假数据

fake_edges <- st_sf(
  data.frame(
    id = c('a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i'),
    weight = c(102.1, 98.3, 201.0, 152.3, 176.4, 108.6, 151.4, 186.3, 191.2), 
    soc = c(-0.1, 0.7, 1.1, 0.2, 0.5, -0.2, 0.4, 0.3, 0.8), 
    geometry = st_sfc(
      st_linestring(rbind(c(1,1), c(1,2))), 
      st_linestring(rbind(c(1,2), c(2,2))), 
      st_linestring(rbind(c(2,2), c(2,3))), 
      st_linestring(rbind(c(1,1), c(2,1))), 
      st_linestring(rbind(c(2,1), c(2,2))), 
      st_linestring(rbind(c(2,2), c(3,2))), 
      st_linestring(rbind(c(1,1), c(1,0))), 
      st_linestring(rbind(c(1,0), c(0,0))), 
      st_linestring(rbind(c(0,0), c(0,1)))
    )
  )
)

从表中创建一个网络,并找到从第一个节点到所有其他节点的最短路径

fake_net <- as_sfnetwork(fake_edges)
fake_paths <- st_network_paths(
  x = fake_net, 
  from = V(fake_net)[1], 
  to = V(fake_net),
  weights = 'weight', 
  type = 'shortest'
)

提取路径中最后一条边的id

idx_numeric <- unlist(lapply(fake_paths[["edge_paths"]], tail, n = 1L))
id <- fake_edges[["id"]][idx_numeric]

对于每条路径,计算路径所有边的 soc 之和

result <- tapply(
  X = fake_edges[["soc"]][unlist(fake_paths[["edge_paths"]])], 
  INDEX = rep(seq_len(nrow(fake_paths)), times = lengths(fake_paths[["edge_paths"]])), 
  FUN = sum
)

用列 id 和 result 创建一个 tibble 对象

my_tbl <- tibble(
  id = id, 
  result = result
)

运行左连接

left_join(fake_edges, my_tbl)
#> Joining, by = "id"
#> Simple feature collection with 9 features and 4 fields
#> Geometry type: LINESTRING
#> Dimension:     XY
#> Bounding box:  xmin: 0 ymin: 0 xmax: 3 ymax: 3
#> CRS:           NA
#>   id weight  soc result              geometry
#> 1  a  102.1 -0.1   -0.1 LINESTRING (1 1, 1 2)
#> 2  b   98.3  0.7    0.6 LINESTRING (1 2, 2 2)
#> 3  c  201.0  1.1    1.7 LINESTRING (2 2, 2 3)
#> 4  d  152.3  0.2    0.2 LINESTRING (1 1, 2 1)
#> 5  e  176.4  0.5     NA LINESTRING (2 1, 2 2)
#> 6  f  108.6 -0.2    0.4 LINESTRING (2 2, 3 2)
#> 7  g  151.4  0.4    0.4 LINESTRING (1 1, 1 0)
#> 8  h  186.3  0.3    0.7 LINESTRING (1 0, 0 0)
#> 9  i  191.2  0.8    1.5 LINESTRING (0 0, 0 1)

我真的不明白算法背后的想法(所以我不确定如何模拟更大的网络),但我认为相同的“算法”在更大的网络上效果很好,你能测试一下吗?


推荐阅读