如何用data.table语法替代循环，按country拼接对应survey字符串？

我有一个data.table，其中country列存在重复值，survey列的每个观测都是唯一字符串。我希望创建一个新变量，将单个country对应的所有survey字符串按行累积拼接起来。目前我已经通过索引和for循环实现了该功能，但想知道有没有更简便的data.table原生语法实现方式。

我的示例代码如下：

library(data.table)

dt <- data.table(country = c("Belgium","Belgium","Bolivia","Brazil","Brazil","Brazil"),
                      survey = c("BE01, BE03, BE04","BE05, BE07, BE11",
                            "BO11, BO13", "BR01, BR02, BR03", "BR05","BR12, BR13"))
dt[, index := seq(1, .N), by = "country"]
for(cntry in unique(dt$country)){
  tmp_ind <- max(dt[country == eval(cntry)]$index)
  dt[country == eval(cntry) & index ==1, all_surveys := survey]
  if(tmp_ind > 1) {
    for(i in 2:tmp_ind){
      dt[country == eval(cntry) & index ==i,
         all_surveys := paste0(dt[country == eval(cntry) & index == i-1, all_surveys], survey)]
    }
  }
}

运行后得到的预期结果：

> dt
   country           survey index                      all_surveys
1: Belgium BE01, BE03, BE04     1                 BE01, BE03, BE04
2: Belgium BE05, BE07, BE11     2 BE01, BE03, BE04BE05, BE07, BE11
3: Bolivia       BO11, BO13     1                       BO11, BO13
4:  Brazil BR01, BR02, BR03     1                 BR01, BR02, BR03
5:  Brazil             BR05     2             BR01, BR02, BR03BR05
6:  Brazil       BR12, BR13     3   BR01, BR02, BR03BR05BR12, BR13

在data.table里可以直接利用分组结合累积拼接的方式，不需要额外创建索引和嵌套循环，一行代码就能达成需求：

library(data.table)

dt <- data.table(country = c("Belgium","Belgium","Bolivia","Brazil","Brazil","Brazil"),
                      survey = c("BE01, BE03, BE04","BE05, BE07, BE11",
                            "BO11, BO13", "BR01, BR02, BR03", "BR05","BR12, BR13"))

# 核心逻辑：按country分组，生成累积拼接的字符串
dt[, all_surveys := Reduce(paste0, survey, accumulate = TRUE), by = country]

运行后得到的结果和你用for循环实现的完全一致：

> dt
   country           survey                      all_surveys
1: Belgium BE01, BE03, BE04                 BE01, BE03, BE04
2: Belgium BE05, BE07, BE11 BE01, BE03, BE04BE05, BE07, BE11
3: Bolivia       BO11, BO13                       BO11, BO13
4:  Brazil BR01, BR02, BR03                 BR01, BR02, BR03
5:  Brazil             BR05             BR01, BR02, BR03BR05
6:  Brazil       BR12, BR13   BR01, BR02, BR03BR05BR12, BR13

代码说明 #

• Reduce(paste0, survey, accumulate = TRUE)：Reduce会依次对当前分组内的survey元素执行paste0拼接，accumulate=TRUE会保留每一步的拼接结果，正好实现按行累积拼接的效果。
• by = country：确保每个国家的拼接独立进行，不会跨组混合内容。

如果需要在拼接的字符串之间添加分隔符（比如逗号加空格），只需把paste0替换为带sep参数的paste即可：

dt[, all_surveys := Reduce(function(x, y) paste(x, y, sep = ", "), survey, accumulate = TRUE), by = country]

内容的提问来源于stack exchange，提问作者BLP92