如何用R的循环或apply生成MARKET与MEAL组合的子数据框？

给定如下R语言示例数据框：

df <- data.frame (MARKET  = c("US", "US", "UK", "UK", "China", "China", "Brazil", "Brazil", "US", "US", "UK", "UK", "China", "China", "Brazil", "Brazil","US", "US", "UK", "UK", "China", "China", "Brazil", "Brazil","US", "US", "UK", "UK", "China", "China", "Brazil", "Brazil","US", "US", "UK", "UK", "China", "China", "Brazil", "Brazil"),
                  MEAL = c("Breakfast", "Lunch", "Dinner", "Breakfast", "Lunch", "Dinner","Breakfast", "Lunch", "Dinner","Breakfast", "Lunch", "Dinner","Breakfast", "Lunch", "Dinner","Breakfast", "Lunch", "Dinner","Breakfast", "Lunch", "Dinner","Breakfast", "Lunch", "Dinner","Breakfast", "Lunch", "Dinner","Breakfast", "Lunch", "Dinner", "Breakfast", "Lunch", "Dinner", "Breakfast", "Lunch", "Dinner", "Breakfast", "Lunch", "Dinner", "Breakfast")
)

需求是生成MARKET与MEAL所有组合对应的独立子数据框（如Brazil_Breakfast、Brazil_Lunch等）。已通过以下代码获取类别向量：

markets <- rownames(table(df$MARKET))
meals <- rownames(table(df$MEAL))

手动子集单个组合的代码可正常运行：

brazil_breakfast <- subset(df, MARKET==markets[1] & MEAL==meals[1])

但编写的嵌套for循环仅生成最后一个组合（US与Lunch）且数据框命名为i_j：

for (i in length(markets)) {
  for (j in length(meals)) {
    i_j <- subset(df, MARKET==markets[i] & MEAL==meals[j]) 
  }
}

解决方案 #

一、修复嵌套for循环 #

原循环的核心问题是for (i in length(markets))仅遍历了最后一个索引（length(markets)返回单个数值），需改为遍历所有索引；同时要动态生成变量名，使用assign()函数将子数据框存入全局环境。

markets <- rownames(table(df$MARKET))
meals <- rownames(table(df$MEAL))

# 修复后的循环
for (i in seq_along(markets)) {
  for (j in seq_along(meals)) {
    # 生成规范的变量名（小写+下划线）
    var_name <- tolower(paste0(markets[i], "_", meals[j]))
    # 子集数据并赋值到全局环境
    assign(var_name, subset(df, MARKET == markets[i] & MEAL == meals[j]))
  }
}

• seq_along(markets)生成从1到length(markets)的序列，比1:length(markets)更安全（避免空向量场景出错）；
• tolower()统一变量名风格，可根据需求调整为其他格式。

二、使用apply系列函数实现 #

通过expand.grid生成所有组合，结合lapply批量处理，最后用list2env转为独立数据框，代码更简洁：

# 生成MARKET与MEAL的所有组合
combos <- expand.grid(MARKET = markets, MEAL = meals, stringsAsFactors = FALSE)

# 批量生成子数据框列表
df_list <- lapply(1:nrow(combos), function(k) {
  subset(df, MARKET == combos$MARKET[k] & MEAL == combos$MEAL[k])
})

# 给列表元素命名
names(df_list) <- tolower(paste0(combos$MARKET, "_", combos$MEAL))

# 将列表转为全局环境中的独立变量
list2env(df_list, envir = .GlobalEnv)

三、更推荐的做法：用列表统一管理 #

实际R项目中，不建议生成大量独立变量污染全局环境，更推荐用列表存储所有子数据框，方便后续批量操作：

# 一行代码生成带命名的子数据框列表
df_list <- split(df, interaction(df$MARKET, df$MEAL, sep = "_"))
# 统一变量名为小写
names(df_list) <- tolower(names(df_list))

# 调用单个子数据框示例
df_list$brazil_breakfast

• split()结合interaction()可快速按组合分组；
• 列表形式便于批量分析、导出等后续操作，代码更简洁易维护。

内容的提问来源于stack exchange，提问作者BDuff