需要向量化字符串上的函数

时间:2019-06-26 20:11:35

标签: r

我有一个30万行数据帧,其中的列是这样的:

   db$performance[1:10]
   [1] "1C1CCCCCCCCCCCCCCCCCCCCCC" "CCCCCCCCCCCCC"             
    "4321CCCCCCCCCCCCCCCCCCCCC" "321CCCCCCCCCCCCCCCCCCCCCC"
    [5] "CCCCCCCCCCCCCC"            "4321CCCCCCCCCCCCC0"  "211CCCCCCCCCCCCCCCCCCCCCC" "BCCCCCCCCC"     [9] "BCCCCCCCCC"                "8"       

我想搜索该列的每一行并计算最后18个(从右到左)字符元素中出现的“ 4”个数字。我拥有的循环解决方案显然很糟糕,因为它非常慢(6分钟或更长时间)。见下文。我如何向量化解决方案(使用apply和/或dplyr?)

谢谢!

substrRight <- function(x, n){
 substr(x, nchar(x)-n, nchar(x))
}

db$NewVar = NA

for (N in 1:nrow(db)){
db$NewVar[N] = str_count( substrRight(db$performance[N],18), "4")
}

1 个答案:

答案 0 :(得分:4)

str_countsubstr已经向量化。因此,直接将功能应用于整个列

library(stringr)
str_count(substrRight(db$performance, 18), "4") 
#[1] 0 0 0 0 0 1 0 0 0 0

它应该足够快。检查更大数据集上的时间

基准

db1 <- db[rep(seq_len(nrow(db)), 1e5),, drop = FALSE]

system.time({
out <- numeric(nrow(db1))
for (i in seq_len(nrow(db1))){
 out[i]= str_count( substrRight(db1$performance[i],18), "4")
}
})
# user  system elapsed 
# 14.699   0.104  14.755 
system.time({
sapply(db1$performance, function(x) str_count( substrRight(x,18), "4") )

})
# user  system elapsed 
# 14.267   0.075  14.299 
system.time({
str_count(substrRight(db1$performance, 18), "4") 

})
# user  system elapsed 
#  0.437   0.016   0.452 

数据

db <- structure(list(performance = c("1C1CCCCCCCCCCCCCCCCCCCCCC", "CCCCCCCCCCCCC", 
"4321CCCCCCCCCCCCCCCCCCCCC", "321CCCCCCCCCCCCCCCCCCCCCC", "CCCCCCCCCCCCCC", 
"4321CCCCCCCCCCCCC0", "211CCCCCCCCCCCCCCCCCCCCCC", "BCCCCCCCCC", 
"BCCCCCCCCC", "8")), class = "data.frame", row.names = c(NA, 
-10L))