如果所有计算都相同，如何按组优化 dplyr 代码答案

【问题标题】：How can I optimize the dplyr code by group if all calculations are the same如果所有计算都相同，如何按组优化 dplyr 代码
【发布时间】：2019-12-09 15:17:49
【问题描述】：

我有以下数据框，它是包含超过 300 万行的更大数据框的子集。

df <- data.frame(Group = c(1,1,1,2,2,3,3,3,2,2,4,4,1,4,1,3,1,3,2,4,2,1,3,2,4),
              SubGroup = c("A","A","C","B","C","A","B","C","C","A","B","C","C","A","B","C","A","A","B","A","C","C","B","B","C"),
              Type = c("Z","Y","Z","X","X","Y","Z","X","Z","Y","Y","Y","X","Y","Z","X","X","Z","Y","X","Z","Z","Y","X","Y"),
              Value = c(4,7,2,8,4,6,1,8,3,6,3,6,3,5,6,7,3,2,7,4,9,4,7,8,1),
              Dup = c(1,1,0,0,0,1,0,1,1,0,1,0,0,1,0,1,0,1,0,1,1,0,1,1,1))

对于每个组，我要计算五个不同的公式。 1) 子组的价值总和 2) SubGroup 和 Dup 的 Value 之和 3) 按子组和类型 Z 的值之和 4) 按子组和类型 Y 的值总和 5) SubGroup 和 X 类型的值之和

这是我目前使用 dplyr 的代码：

result <- df %>%
 group_by(Group) %>%
 summarise(SubGroup.A.Total = sum(Value[SubGroup=="A"]),
        SubGroup.A.Dup = sum(Value[SubGroup=="A" & Dup==1]),
        SubGroup.A.TypeZ = sum(Value[SubGroup=="A" & Type=="Z"]),
        SubGroup.A.TypeY = sum(Value[SubGroup=="A" & Type=="Y"]),
        SubGroup.A.TypeX = sum(Value[SubGroup=="A" & Type=="X"]),
        SubGroup.B.Total = sum(Value[SubGroup=="B"]),
        SubGroup.B.Dup = sum(Value[SubGroup=="B" & Dup==1]),
        SubGroup.B.TypeZ = sum(Value[SubGroup=="B" & Type=="Z"]),
        SubGroup.B.TypeY = sum(Value[SubGroup=="B" & Type=="Y"]),
        SubGroup.B.TypeX = sum(Value[SubGroup=="B" & Type=="X"]),
        SubGroup.C.Total = sum(Value[SubGroup=="C"]),
        SubGroup.C.Dup = sum(Value[SubGroup=="C" & Dup==1]),
        SubGroup.C.TypeZ = sum(Value[SubGroup=="C" & Type=="Z"]),
        SubGroup.C.TypeY = sum(Value[SubGroup=="C" & Type=="Y"]),
        SubGroup.C.TypeX = sum(Value[SubGroup=="C" & Type=="X"]))

但是，我真的很想知道从运行时间和行数的角度来看，是否有更有效的方法？由于我本质上是按 SubGroup 循环相同的五个公式，我想有一种更简单的方法来编写它。

【问题讨论】：

标签： r optimization dplyr summarize

【解决方案1】：

此答案使用pivot_wider，目前仅在tidyr 的开发版中可用。

您将按SubGroup 对要计算的所有变量进行分组，因此值得将其添加到group_by。对于某些变量，您还按Type 分组，所以我想您可以按Group, SubGroup, Type 进行单独分组并将这些变量重新加入，但尚不清楚这是否值得。

坚持group_by(Group, SubGroup)，你可以这样做：

df %>%
    group_by(Group, SubGroup) %>%
    summarize(
        Total = sum(Value),
        Dup = sum(Value[Dup == 1]),
        X = sum(Value[Type == "X"]),
        Y = sum(Value[Type == "Y"]),
        Z = sum(Value[Type == "Z"])
    ) %>%
    pivot_wider(
        names_from = SubGroup,
        values_from = c(-Group, -SubGroup)
    )

【讨论】：

【解决方案2】：

类似于@Marius 的回答：

do.call(
  rbind,
  by(df, list(df$Group, df$SubGroup), FUN = function(x) {
    within(x, {
      value_sum <- sum(Value)
      value_sum_dup1 <- sum(Value[Dup == 1])
      value_sum_dup0 <- sum(Value[Dup == 0])
      value_z <- sum(Value[Type == "Z"])
      value_y <- sum(Value[Type == "Y"])
      value_x <- sum(Value[Type == "Z"])
    })[1, ]
  })
)

【讨论】：