在 for 循环中优化 add_trace()? [英] Optimizing add_trace() in a for loop?

问题描述

我在 for 循环中使用 add_trace() 函数在 plotly 的 scatter3d 模式下为 3d 网络图创建线条.每个 add_trace 在网络中的两个节点之间绘制一条单独的线.该方法是有效的，但是对于大量循环，各个循环的速度似乎很快就会减慢.

示例数据可以在这里下载:

此问题的可重复数据

为方便起见，这里是这个问题的数据.OP 需要从 github 下载一个 .gml 文件，并安装 library(igraph) 以将数据处理到这些文件中.

es <- 结构(列表(V1 = c(1, 1, 2, 1, 2, 3, 1, 1, 1, 5, 6, 1, 2, 3, 4, 1, 3, 3, 1, 5, 6, 1, 1,4, 1, 2, 3, 4, 6, 7, 1, 2, 1, 2,1, 2, 24, 25, 3, 24, 25, 3, 24, 27, 2, 9, 1, 25, 26, 29, 3, 9, 15, 16, 19, 21, 23, 24, 30,31, 32, 9, 10, 14, 15, 16, 19, 20,21, 23, 24, 27, 28, 29, 30, 31, 32, 33),V2 = c(2, 3, 3, 4, 4, 4, 5, 6, 7, 7, 7, 8, 8, 8, 8, 9, 9, 10, 11, 11, 11, 12, 13,13,14, 14, 14, 14, 17, 17, 18, 18, 20, 20, 22, 22, 26, 26, 28, 28, 28, 29, 30, 30, 31, 31, 32, 32, 32,32, 33, 33, 33, 33, 33, 33,33, 33, 33, 33, 33, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34)),.Names = c("V1", "V2"), row.names = c(NA, -78L), class = "data.frame")theta <- seq(0,2,length.out=35)[1:34]Xn <- cospi(θ)Yn <- sinpi(θ)Nv <- NROW(Xn)Ne <- NROW(es)vs <- data.frame(label = as.character(1:Nv))

I'm using the add_trace() function in a for loop to create lines for a 3d network graph in plotly's scatter3d mode. Each add_trace draws an individual line between two nodes in the network. The method is working, but with large number of loops, the speed of the individual loops seems to be slowing down very quickly.

Example data can be downloaded here: https://gist.github.com/pravj/9168fe52823c1702a07b

library(igraph)
library(plotly)

G <- read.graph("karate.gml", format = c("gml"))
L <- layout.circle(G)

vs <- V(G)
es <- as.data.frame(get.edgelist(G))

Nv <- length(vs)
Ne <- length(es[1]$V1)

Xn <- L[,1]
Yn <- L[,2]

network <- plot_ly(type = "scatter3d", x = Xn, y = Yn, z = rep(0, Ne), mode = "markers", text = vs$label, hoverinfo = "text", showlegend = F)

for(i in 1:Ne) {
  v0 <- es[i,]$V1
  v1 <- es[i,]$V2

  x0 <-  Xn[v0]
  y0 <-  Yn[v0]
  x1 <-  Xn[v1]
  y1 <-  Yn[v1]

  df <-  data.frame(x = c(x0, x1), y = c(y0, y1), z = c(0, 0))
  network <- add_trace(network, data = df, x = x, y = y, z = z, type = "scatter3d", mode = "lines", showlegend = F, 
                       marker = list(color = '#030303'), line = list(width = 0.5))
}

This example is fairly quick, but when I include a few hundred edges or more, the execution of the individual loops start to slow down radically. I tried different optimization methods (vectorisation etc), but there seems to be no working around the slowness of the add_trace function itself.

Any suggestions?

解决方案

The most efficient way to add many line segments in plotly is not as a separate trace each, but to use only a single trace that contains all the line segments. You can do this by constructing a data frame with the x,y coordinates of each node to be connected, interspersed with NA's between each line segment. Then use connectgaps=FALSE to break the trace into separate segments at each NA. You can see another example of this approach, applied to spaghetti plots in this answer.

es$breaks <- NA
lines <- data.frame(node=as.vector(t(es)), x=NA, y=NA, z=0)
lines[which(!is.na(lines$node)),]$x <- Xn[lines[which(!is.na(lines$node)),]$node]
lines[which(!is.na(lines$node)),]$y <- Yn[lines[which(!is.na(lines$node)),]$node]

network <- plot_ly(type = "scatter3d", x = Xn, y = Yn, z = rep(0, Ne), 
                   mode = "markers", text = vs$label, hoverinfo = "text", 
                   showlegend = F) %>% 
  add_trace(data=lines, x=x, y=y, z=z, showlegend = FALSE,
                      type = 'scatter3d', mode = 'lines+markers',
                      marker = list(color = '#030303'), line = list(width = 0.5),
                      connectgaps=FALSE)

Reproducible data for this question

For convenience, here are the data for this question. The OP required downloading a .gml file from github, and installing library(igraph) to process the data into these.

es <- structure(list(
  V1 = c(1, 1, 2, 1, 2, 3, 1, 1, 1, 5, 6, 1, 2, 3, 4, 1, 3, 3, 1, 5, 6, 1, 1, 4, 1, 2, 3, 4, 6, 7, 1, 2, 1, 2, 
    1, 2, 24, 25, 3, 24, 25, 3, 24, 27, 2, 9, 1, 25, 26, 29, 3, 9, 15, 16, 19, 21, 23, 24, 30, 31, 32, 9, 10, 14, 15, 16, 19, 20, 
    21, 23, 24, 27, 28, 29, 30, 31, 32, 33), 
  V2 = c(2, 3, 3, 4, 4, 4, 5, 6, 7, 7, 7, 8, 8, 8, 8, 9, 9, 10, 11, 11, 11, 12, 13, 13, 
    14, 14, 14, 14, 17, 17, 18, 18, 20, 20, 22, 22, 26, 26, 28, 28, 28, 29, 30, 30, 31, 31, 32, 32, 32, 32, 33, 33, 33, 33, 33, 33, 
    33, 33, 33, 33, 33, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34, 34)), 
  .Names = c("V1", "V2"), row.names = c(NA, -78L), class = "data.frame")

theta <- seq(0,2,length.out=35)[1:34]
Xn <- cospi(theta)
Yn <- sinpi(theta)

Nv <- NROW(Xn)
Ne <- NROW(es)
vs <- data.frame(label = as.character(1:Nv))

这篇关于在 for 循环中优化 add_trace()?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！