我有一个聚合操作,该操作对R中多边形的点进行计数,该操作当前使用plyr :: ddply()作为主要功能,在这里我需要按2个变量进行分组:dayofweek和hour。它非常慢,所以我想用一个更快的函数代替它,例如data.table :: package中的东西。
该操作的主要目的是获取点 df 的数据框,并使用sf ::包中的st_intersects方法计算网格中与多边形相交的点数.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)
# 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()
这在我的机器上花费了大约4秒钟,但是我必须执行该过程数百次才能真正实现。
我尝试从data.table开始,但发现翻译困难。这是我为此data.table尝试所拥有的唯一(不起作用的)代码:
test4 <- grid.sf[, .(test = myf()), by = key(grid.sf)]
因此,如果有一种方法可以将此ddply转换为data.table操作,则由于data.table快得多,所以我会很高兴。
谢谢!