如何避免函数中含有多个条件的for循环

I have a 25 years data set that looks similar to the following:

我有一个25年的数据集，看起来类似如下:

        date name        value tag
1 2014-12-01    f -0.338578654  12
2 2014-12-01    a  0.323379254   4
3 2014-12-01    f  0.004163806   9
4 2014-12-01    f  1.365219477   2
5 2014-12-01    l -1.225602543   7
6 2014-12-01    d -0.308544089   9

This is how to replicate it:

这就是如何复制它:

set.seed(9)
date <- rep(seq(as.Date("1990-01-01"), as.Date("2015-01-1"), by="months"), each=50)
N <- length(date)
name <- sample(letters, N, replace=T)
value <- rnorm(N)
tag <- sample(c(1:50), N, replace=T)
mydata <- data.frame(date, name, value, tag)
head(mydata)

I would like to create a new matrix that stores values that satisfy multiple criteria. For instance, the sum of values that have a name j and a tag i. I use two for-loops and the which() function to filter out the correct values. Like this:

我想创建一个新的矩阵来存储满足多个条件的值。例如，具有名称为j和标记为i的值的和，我使用了两个For循环和一个which()函数来过滤出正确的值。是这样的:

S <- matrix(data=NA, nrow=length(unique(mydata$tag)), ncol=length(unique(mydata$name)))
for(i in 1:nrow(S)){
  for (j in 1:ncol(S)){
    foo <- which(mydata$tag == unique(mydata$tag)[i] & mydata$name == unique(mydata$name)[j])
    S[i,j] <- sum(mydata$value[foo])
  }
}

This is ok for small data sets, but too slow for larger ones. Is it possible to avoid the for-loops or somehow speed up the process?

这对于小数据集来说是可以的，但是对于大数据集来说太慢了。是否有可能避免for循环或以某种方式加快进程?

1 个解决方案

#1

You can use dcast from package reshape2, with a custom function to sum your values:

您可以使用包reshape2中的dcast，并使用自定义函数来求和您的值:

library(reshape2)
dcast(mydata, name~tag, value.var='value', fun.aggregate=sum)

Or simply xtabs, base R:

或者简单的xtabs，基数为R:

xtabs(value~name+tag, mydata)

Some benchmark:

一些指标:

funcPer = function(){
    S <- matrix(data=NA, nrow=length(unique(mydata$tag)), ncol=length(unique(mydata$name)))
    for(i in 1:nrow(S)){
      for (j in 1:ncol(S)){
        foo <- which(mydata$tag == unique(mydata$tag)[i] & mydata$name == unique(mydata$name)[j])
        S[i,j] <- sum(mydata$value[foo])
      }
    }
}

colonel1 = function() dcast(mydata, name~tag, value.var='value', fun.aggregate=sum)

colonel2 = function() xtabs(value~name+tag, mydata)

#> system.time(colonel1())
#  user  system elapsed 
#   0.01    0.00    0.01 
#> system.time(colonel2())
#   user  system elapsed 
#   0.05    0.00    0.05 
#> system.time(funcPer())
#   user  system elapsed 
#   4.67    0.00    4.82

#1