如何在R数据帧中对1和0分隔的空格字符串进行求和

Question

我有一个数据框，其列有一串字符串，如'0 1 0 0 0 1 1 0'

它们的长度都相同，但包含不同数量的1和0。我如何在r？

中总结它们

Answer 1

如果它们是字符串，您可以使用scan创建数字向量，只需sum：

sum(scan(text = "0 1 0 0 0 1 1 0", quiet = TRUE))
# [1] 3

如果是列，您仍然可以使用scan。方法可能类似于：

rowSums(matrix(scan(text = as.character(df$V1)), ncol = 8, byrow = TRUE))

将“df$V1”替换为您的数据实际所在的列，并将“ncol = 8”替换为您应该拥有的多列。

vapply + strsplit也是不错的选择。这与@karen's approach with sapply类似，但不是两次循环遍历，只需执行一次：

vapply(strsplit(as.character(x), " ", TRUE), function(y)
       sum(as.integer(y)), integer(1L))

但更快的选择是使用“data.table”中的fread或“iotools”中的dstrstplit来读取数据，然后使用rowSums()

library(data.table)
rowSums(fread(paste(x, collapse = "\n")))

library(iotools)
rowSums(dstrsplit(x, sep = " ", col_types = rep("integer", 10)))

以下是一些基准测试。首先，要测试的样本数据和函数：

set.seed(1)
mydf <- data.frame(V1 = replicate(10000, paste(sample(c(0, 1), 10, TRUE), collapse = " ")))

library(data.table) # For fun_fread()
library(iotools)    # For fun_iotools()

fun_scan <- function(x) {
  nc <- length(scan(text = as.character(x[1]), quiet = TRUE))
  rowSums(matrix(scan(text = as.character(x), quiet = TRUE), ncol = nc, byrow = TRUE))
}

fun_vapply <- function(x) {
  vapply(strsplit(as.character(x), " ", TRUE), function(y) sum(as.integer(y)), integer(1L))
}

fun_sapply <- function(x) {
  sapply(as.character(x), function(col) { sum(sapply(strsplit(col, ' '), as.integer))})
}

fun_fread <- function(x) rowSums(fread(paste(x, collapse = "\n")))

fun_iotools <- function(x) {
  nc <- length(scan(text = as.character(x[1]), quiet = TRUE))
  rowSums(dstrsplit(as.character(x), sep = " ", col_types = rep("integer", nc)))
} 

检查一切是否相等：

all.equal(fun_scan(mydf$V1), fun_vapply(mydf$V1))
# [1] TRUE

all.equal(unname(fun_sapply(mydf$V1)), fun_vapply(mydf$V1))
# [1] TRUE

all.equal(fun_fread(mydf$V1), fun_vapply(mydf$V1))
# [1] TRUE

all.equal(fun_fread(mydf$V1), fun_iotools(mydf$V1))
# [1] TRUE

现在，基准：

library(microbenchmark)
res <- microbenchmark(fun_scan(mydf$V1), fun_vapply(mydf$V1), 
                      fun_sapply(mydf$V1), fun_fread(mydf$V1), fun_iotools(mydf$V1))
# Unit: milliseconds
#                  expr        min         lq       mean     median         uq        max neval
#     fun_scan(mydf$V1)  52.071714  54.473772  57.257971  55.869307  58.200922  74.248057   100
#   fun_vapply(mydf$V1)  20.477943  22.079925  24.796633  23.613694  26.205056  46.857130   100
#   fun_sapply(mydf$V1) 267.655633 281.656375 308.750804 296.128511 319.318740 449.806432   100
#    fun_fread(mydf$V1)   8.112898   8.391891   9.154349   8.643196   9.434840  16.967090   100
#  fun_iotools(mydf$V1)   4.436515   4.608810   4.967436   4.751517   5.114486   8.690872   100

autoplot(res)

Answer 2

我愿意：

strTest <- '0 1 0 0 0 1 1 0'
sum(sapply(strsplit(strTest, ' '), as.integer))

考虑到它是一个名为test_col的列，那么：

sapply(test_col, function(col) { sum(sapply(strsplit(col, ' '), as.integer))})