r - 如何用 data.table 替换 plyr::ddply

Question

我有一个聚合操作，它计算 R 中当前用plyr::ddply()作主函数的多边形中的点，我需要按 2 个变量分组：dayofweek和hour. 它很慢，所以我想用更快的功能替换它，比如data.table包中的东西。

代表

创建数据框

该操作的主要目标是获取点的数据框 df，并使用包中的st_intersects()方法sf计算有多少点与多边形相交grid.sf。

创建 DF 对象

library(sf)
library(tidyverse)
library(plyr)

df <- data.frame(X = seq(1,100,1), 
                 dayofweek = rep(c("Sun", "Mon", "Tues", "Wed", "Thur"), 20),
                 hour = sample(seq(0, 23, 1),  100, replace = T),
                 lat = sample(seq(37.1234, 37.2345, 0.001),  100, replace = T),
                 lon = sample(seq(-122.5432, -122.4111, 0.001),  100, replace = T)
)


projcrs <- "+proj=longlat +datum=WGS84 +no_defs +ellps=WGS84 +towgs84=0,0,0"

df <- st_as_sf(x = df,                         
               coords = c("lon", "lat"),
               crs = projcrs)

创建 grid.sf 对象

# Function to create the grid we need
buildBaseGrid <- function(x) {
  
  # create a 500m tesseract over these points
  g <- st_make_grid(x, cellsize = 0.005)
  
  # plot to make sure
  # ggplot() +
  #   geom_sf(data = df.sf, size = 3) +
  #   geom_sf(data = g, alpha = 0)
  # 
  # ggplot() +
  #   geom_sf(data = g, alpha = 0)
  grid.sf <- st_sf(g) 
  # Need to expand by day of week
  days <- c('SU', 'MO', 'TU', 'WE', 'TH', 'FR', 'SA')
  hours <- c('0', '1', '2', '3', '4', '5', '6', '7', '8', '9', '10', '11', '12',
             '13', '14', '15', '16', '17', '18', '19', '20', '21', '22', '23')
  grid.sf <- expand.grid(g, days, hours)
  grid.sf$id <- 1:nrow(grid.sf)
  
  #### Clean up home grid
  
  # Calc centroid
  grid.sf <- grid.sf %>%
    dplyr::mutate(center = sf::st_centroid(grid.sf$Var1))
  
  # Parse out lat and lon
  grid.sf <- grid.sf %>%
    dplyr::mutate(lonn = sf::st_coordinates(grid.sf$center)[,1]) %>%
    dplyr::mutate(latt = sf::st_coordinates(grid.sf$center)[,2])
  
  # Create primary key field
  grid.sf <- grid.sf %>%
    dplyr::mutate(pkey = paste0(lonn,";",latt,";",Var2,";",Var3))
  
  
  grid.sf <- st_as_sf(grid.sf) 
  
  return(grid.sf)
  
  
}

# Now build the grid.sf object
grid.sf <- buildBaseGrid(df)

plyr::ddply 的当前操作

# Create function to use in operation
myf <- function(x) {
  x <- as.data.frame(x)
  df <- df %>% dplyr::filter(dayofweek %in% x$Var2)
  df <- df %>% dplyr::filter(hour %in% x$Var3)
  x$count <- sf::st_intersects(x$Var1, df) %>% lengths()
  x %>%
    data.frame(x)
}

# Do the operation
output <-  plyr::ddply(grid.sf, .(Var2, Var3), .fun = myf, .parallel = F) %>% as.data.frame()

这在我的机器上大约需要 4 秒，但我必须这样做数百次，所以它真的加起来了。

我尝试从 data.table 开始，但发现翻译很困难。这是我进行此data.table()尝试的唯一（无效）代码：

test4 <- grid.sf[, .(test = myf()), by = key(grid.sf)]

所以，如果有办法将其转化ddply为data.table操作，我会被粉红色挠痒痒的，因为data.table速度要快得多。

谢谢！

标签： rdata.tablegeospatialplyrsf