如何在 R 中使用 foreach（） 解决“外部指针无效”

问：

我正在编写一个代码，用于计算与每个人口像素单元格的给定距离内的公园面积。这两个数据集都是 30m 栅格，我从中提取类和 xy 坐标（以米为单位）以创建数据帧。由于数据集非常庞大，我正在尝试逐个县进行计算。

人口数据的 TIFF 大致代表每个县，因此我在读取它们的同一调用中使用它们来裁剪公园栅格。我的代码运行良好，但我一直在努力使它成为一个并行过程，因为在整个国家/地区执行此操作的处理时间约为一整个工作日。我已经将我的原始代码改编为使用 foreach（），它比原始代码更有效地使用我的机器。

当我手动运行每个步骤时，此代码起作用，但在我尝试并行运行时抛出以下错误：“{ 中的错误：任务 1 失败 - ”i 在参数中：'Count = dist_Euclidean（...）'。由错误引起：！在为函数“as.data.frame”选择方法时计算参数“x”时出错：外部指针无效” 需要强调的是，如果我手动设置“code”和“i”并在不运行 foreach（） 的情况下单击它，则此代码会为我提供我想要的确切输出。问题出在并行化上。

**注意：我知道提供您使用的数据是最佳做法，但这些数据集非常大。“县”和“州”是美国人口普查局的形状文件，“公园”是我对全国公园区域的栅格（30m 像素）。“rasterlist”提取我拥有的每个群体 TIFF 的文件名。我认为我无法提供这些数据进行测试，所以希望这可以是一个理论练习？

pkgList <- c("parallel", "parallelly","doParallel","foreach","terra","sf","dplyr","tidyr","stringi","stringr")
for(package.i in pkgList) {
  suppressPackageStartupMessages(library(package.i, character.only = TRUE))} ; rm(package.i)

rasterlist <- as.list(gsub("._","",list.files(path = "tifs", pattern = '.tif$', all.files = F, full.names = T)))
states <- st_read("StateLines/cb_2018_us_state_500k.shp")
counties <- st_read("CountyLines/cb_2018_us_county_500k.shp") %>% 
  mutate(State = stri_replace_all_fixed(STATEFP, states$STATEFP, states$NAME, vectorize_all = FALSE),
         code_st = paste(STATEFP,"-", sep = ""),
         State = gsub(" ", "", State), StateCode = paste(State, "-", sep = ""),
         NAME = case_when(NAME == "Miami-Dade" ~ "MiamiDade", TRUE ~ NAME)) %>% #stupid Miami-Dade
  filter(!State %in% c("Alaska","Hawaii","CommonwealthoftheNorthernMarianaIslands","UnitedStatesVirginIslands","AmericanSamoa","Guam","PuertoRico")) #I only have pop data for the continental US at the moment
parks <- rast("ParksProject.tif")

dist_Euclidean <- function(x, y, dfrm, accDist) {
  count <- 0
  dfrm1 <- data.frame(x=x,y=y)
  for (i in 1:nrow(dfrm1)){
    hold <- dfrm1[i,]
    count <- append(count, length(which(sqrt((dfrm$x - hold$x)^2 + (dfrm$y - hold$y)^2)<accDist)))}
  return(count[-1])}

accessDist <- 500 ###Distance used to define "access", in meters

cl <- makeCluster(availableCores(omit = 1, constraints = "connections", logical = FALSE), type = "PSOCK")
registerDoParallel(cl)

  for (code in unique(counties$code_st)) {
    df.state <- as.character(rasterlist[str_detect(rasterlist, code)])
    counties_temp <- st_transform(counties %>% filter(code_st == code), st_crs(parks))
    parks_state <- crop(parks, extend(ext(counties_temp), accessDist/30+5))
write.csv(  
    foreach(i = 1:length(df.state), .packages = pkgList, .combine = "rbind") %dopar% {
      temp <- as.data.frame(rast(df.state[i]), xy = TRUE) %>% 
        mutate(File = paste(df.state[i]), 
               State = stri_replace_all_fixed(File, counties_temp$code_st, counties_temp$StateCode, vectorize_all = FALSE),
               State = stri_replace_all_fixed(State, counties_temp$COUNTYFP, counties_temp$NAME, vectorize_all = FALSE)) %>%
        separate_wider_delim(State, names = c("Trash1","Trash2","State","County"), delim = c("-")) %>% 
        mutate(County = str_remove(County, ".tif")) %>% select(-c("Trash1","Trash2")) %>% 
        rename("Pop" = starts_with("neon-"))%>% mutate(Count = dist_Euclidean(x, y, as.data.frame(crop(parks_state, extend(ext(rast(df.state[i])),accessDist/30+5)), xy = TRUE), accDist = accessDist))
    }, (paste(unique(counties_temp$State), "Parks.csv", sep = "")))
rm(df.state, counties_temp, parks_state)
gc()
  }
stopCluster(cl)

重要的一点：在我尝试将 dist_Euclidean 函数添加到其中之前，它与 foreach（） 循环并行运行。如果我下降%>%突变（计数= dist_Euclidean...accessDist）），这有效 - 它只是创建人口数据的数据帧。问题似乎出在公园数据帧创建中，但只有当我尝试将其作为并行过程进行时。我尝试使用 foreach 中的 .export 调用手动将parks_state栅格添加到每个并行化中，但这并不能解决问题。

这是怎么回事？

编辑：：添加回溯输出

> traceback()
9: stop(simpleError(msg, call = expr))
8: e$fun(obj, substitute(ex), parent.frame(), e$data)
7: foreach(i = 1:length(df.state), .packages = pkgList, .combine = "rbind") %dopar% 
       {
           temp <- as.data.frame(rast(df.state[i]), xy = TRUE) %>% 
               mutate(File = paste(df.state[i]), State = stri_replace_all_fixed(File, 
                   counties_temp$code_st, counties_temp$StateCode, 
                   vectorize_all = FALSE), State = stri_replace_all_fixed(State, 
                   counties_temp$COUNTYFP, counties_temp$NAME, vectorize_all = FALSE)) %>% 
               separate_wider_delim(State, names = c("Trash1", "Trash2", 
                   "State", "County"), delim = c("-")) %>% mutate(County = str_remove(County, 
               ".tif")) %>% select(-c("Trash1", "Trash2")) %>% rename(Pop = starts_with("neon-")) %>% 
               mutate(Count = dist_Euclidean(x, y, as.data.frame(crop(parks_state, 
                   extend(ext(rast(df.state[i])), accessDist/30 + 
                     5)), xy = TRUE), accDist = accessDist))
       }
6: is.data.frame(x)
5: utils::write.table(foreach(i = 1:length(df.state), .packages = pkgList, 
       .combine = "rbind") %dopar% {
       temp <- as.data.frame(rast(df.state[i]), xy = TRUE) %>% mutate(File = paste(df.state[i]), 
           State = stri_replace_all_fixed(File, counties_temp$code_st, 
               counties_temp$StateCode, vectorize_all = FALSE), 
           State = stri_replace_all_fixed(State, counties_temp$COUNTYFP, 
               counties_temp$NAME, vectorize_all = FALSE)) %>% separate_wider_delim(State, 
           names = c("Trash1", "Trash2", "State", "County"), delim = c("-")) %>% 
           mutate(County = str_remove(County, ".tif")) %>% select(-c("Trash1", 
           "Trash2")) %>% rename(Pop = starts_with("neon-")) %>% 
           mutate(Count = dist_Euclidean(x, y, as.data.frame(crop(parks_state, 
               extend(ext(rast(df.state[i])), accessDist/30 + 5)), 
               xy = TRUE), accDist = accessDist))
   }, (paste(unique(counties_temp$State), "Parks.csv", sep = "")), 
       col.names = NA, sep = ",", dec = ".", qmethod = "double")
4: eval(expr, p)
3: eval(expr, p)
2: eval.parent(Call)
1: write.csv(foreach(i = 1:length(df.state), .packages = pkgList, 
       .combine = "rbind") %dopar% {
       temp <- as.data.frame(rast(df.state[i]), xy = TRUE) %>% mutate(File = paste(df.state[i]), 
           State = stri_replace_all_fixed(File, counties_temp$code_st, 
               counties_temp$StateCode, vectorize_all = FALSE), 
           State = stri_replace_all_fixed(State, counties_temp$COUNTYFP, 
               counties_temp$NAME, vectorize_all = FALSE)) %>% separate_wider_delim(State, 
           names = c("Trash1", "Trash2", "State", "County"), delim = c("-")) %>% 
           mutate(County = str_remove(County, ".tif")) %>% select(-c("Trash1", 
           "Trash2")) %>% rename(Pop = starts_with("neon-")) %>% 
           mutate(Count = dist_Euclidean(x, y, as.data.frame(crop(parks_state, 
               extend(ext(rast(df.state[i])), accessDist/30 + 5)), 
               xy = TRUE), accDist = accessDist))
   }, (paste(unique(counties_temp$State), "Parks.csv", sep = "")))