优化 for 循环答案

【问题标题】：Optimize a for loop优化 for 循环
【发布时间】：2020-04-14 18:13:43
【问题描述】：

df_data 是一个包含 40 000 行和 30 列的数据框，在 for 循环中我尝试添加许多列（大约 600 列），它可以工作，但速度非常慢，我尝试使用不同的替代方案，例如 apply 但我不能达到同样的结果。有没有办法同时添加一行的所有值？

    df_i_final <- data.frame()
    start<- Sys.time()

    for (i in 1:nrow(df_data)) {

      row <- df_data[i, ]

      varname_1 <- paste("1",row["CATEGORIE"],row["RANK"],row["MD"],"a", sep="_")
      varname_2 <- paste("2",row["CATEGORIE"],row["RANK"],row["MD"],"b", sep="_")
      varname_3 <- paste("3",row["CATEGORIE"],row["RANK"],row["MD"],"c", sep="_")
      varname_4 <- paste("4",row["CATEGORIE"],row["RANK"],row["MD"],"d", sep="_")
      varname_5 <- paste("5",row["CATEGORIE"],row["RANK"],row["MD"],"e", sep="_")
      varname_6 <- paste("aa",row["CATEGORIE"],row["RANK"],row["MD"],"fg", sep="_")
      varname_7 <- paste("aa",row["CATEGORIE"],row["RANK"],row["MD"],"fg", sep="_")
      varname_8 <- paste("aa",row["CATEGORIE"],row["RANK"],row["MD"],"fg", sep="_")
      varname_9 <- paste("aa",row["CATEGORIE"],row["RANK"],row["MD"],"fg", sep="_")

      if (row["VAL1"] > 0 | row["VAL2"] > 0 | row["VAL3"] > 0 | row["VAL4"] > 0 | row["VAL5"] > 0 | row["VAL6"] > 0 | row["VAL7"] > 0 )
      {
      df_i_final[i, "IDENT"] <- row["IDENT"]
      if (row["VAL1"] > 0) df_i_final[i, varname_1] <- row["VAL1"]
      if (row["VAL2"] > 0) df_i_final[i, varname_2] <- row["VAL2"] 
      if (row["VAL3"] > 0) df_i_final[i, varname_3] <- row["VAL3"]
      if (row["VAL4"] > 0) df_i_final[i, varname_4] <- row["VAL4"]
      if (row["VAL5"] > 0) df_i_final[i, varname_5] <- row["VAL5"]
      if (row["VAL6"] > 0) df_i_final[i, varname_6] <- row["VAL6"]
      if (row["VAL7"] > 0) df_i_final[i, varname_7] <- row["VAL7"]
      }
    }

    process_time<- Sys.time() - start
    print(format(process_time))

编辑：我添加了一些示例，并尝试使用您的代码来创建列，但是当它设置值时，它应该只更新具有相同类别名称的列...

df2 <- data.frame(ID = c("1100455", "1100455", "1100455", "1100455", "1100455", "1100464", "1100464"),
                  CATEGORIE = c("10110", "10160", "10604", "11220", "90310", "10110","10140"),
                  RANK =  c("1", "1", "1", "1", "0" ,"1", "1"),
                  MD =  c("0", "0", "0", "3", "4", "0", "0" ),
                  PROD3 = c(2345.00,1114.58,501.40,0.00,0.00,2720.00,636.80),
                  VALUE3 = c(540.00,0.00,0.00,0.00,0.00,0.00,0.00),
                  AREA3 = c(563.76,0.00,17.35,0.00,0.00,0.00,0.00),
                  LONG3 = c(4100,2100,1740,265,0,3978,940)
)


nm1 <-c("PROD3")
nm1
i1 <- Reduce(`|`, lapply(df2[nm1], `>`, 0))
newvars <- paste("aa",df2[["CATEGORIE"]],df2[["RANK"]],df2[["MD"]],"ta", sep="_")
newvars <- unique(newvars)
newvars
df2[newvars] <- NA
df2[i1, newvars] <- df2[i1, nm1]
df2

这里所有名称类别不同的列都更新了，它必须只有与类别值匹配的列（所以这里只有aa_10110_1_0_ta）

       ID CATEGORIE RANK MD   PROD3 VALUE3  AREA3 LONG3 aa_10110_1_0_ta aa_10160_1_0_ta aa_10604_1_0_ta aa_11220_1_3_ta aa_90310_0_4_ta aa_10140_1_0_ta
1 1100455     10110    1  0 2345.00    540 563.76  4100         2345.00         2345.00         2345.00         2345.00         2345.00         2345.00

【问题讨论】：

你能用 dput 添加一个可重现的小例子吗
我编辑了我的第一篇文章，但我认为我正在寻找的是一个传播函数

标签： r dataframe for-loop optimization dplyr

【解决方案1】：

我们可以将其矢量化。用lapply 遍历感兴趣的列，检查它是否大于0，Reduce 到| 的单个逻辑向量，使用它来创建具有该列中相应值的新列

nm1 <- paste0('VAL', 1:7)
i1 <- Reduce(`|`, lapply(df_i_final[nm1], `>`, 0))
newvars <- paste(seq_len(nrow(df_i_final)), 
     df_i_final[["CATEGORIE"]],df_i_final[["RANK"]],df_i_final[["MD"]],"a", sep="_")
df_i_final[newvars] <- NA
df_i_final[i1, newvars] <- df_i_final[i1, nm1]

【讨论】：