jay*_*.sf 6 variables r function lapply
我想将具有不同值的多个变量的函数应用于列表.我知道如何使用一个变化的变量来做到这一点
sapply(c(1:10), function(x) x * 2)
# [1] 2 4 6 8 10 12 14 16 18 20
Run Code Online (Sandbox Code Playgroud)
但不是两个.我首先手动向你展示我想要的东西(实际上我使用 lapply()但sapply()在SO中更具概要性):
# manual
a <- sapply(c(1:10), function(x, y=2) x * y)
b <- sapply(c(1:10), function(x, y=3) x * y)
c <- sapply(c(1:10), function(x, y=4) x * y)
c(a, b, c)
# [1] 2 4 6 8 10 12 14 16 18 20 3 6 9 12 15 18 21 24 27 30 4 8 12
# [24] 16 20 24 28 32 36 40
Run Code Online (Sandbox Code Playgroud)
这是我尝试,我尝试定义都x和y.
# attempt
X <- list(x = 1:10, y = 2:4)
sapply(c(1:10, 2:4), function(x, y) x * y)
# Error in FUN(X[[i]], ...) : argument "y" is missing, with no default
Run Code Online (Sandbox Code Playgroud)
解决方案的基准
library(microbenchmark)
microbenchmark(sapply = as.vector(sapply(1:10, function(x, y) x * y, 2:4)),
mapply = mapply( FUN = function(x, y) x * y, 1:10, rep( x = 2:4, each = 10)),
sapply2 = as.vector(sapply(1:10, function(y) sapply(2:4, function(x) x * y))),
outer = c(outer(1:10, 2:4, function(x, y) x * y)))
# Unit: microseconds
# expr min lq mean median uq max neval
# sapply 34.212 36.3500 62.44864 39.1295 41.9090 2304.542 100
# mapply 62.008 65.8570 87.82891 70.3470 76.5480 1283.342 100
# sapply2 196.714 203.9835 262.09990 223.6550 232.2080 3344.129 100
# outer 7.698 10.4775 13.02223 12.4020 13.4715 53.883 100
Run Code Online (Sandbox Code Playgroud)
mapply() 将函数应用于多个列表或矢量参数.
rep()也用于重复值2,3和4.在each参数中指定10 ,rep()重复x10次的每个元素.
这是必要的,因为mapply()- 1:10中的第一个参数长度为10.
# supply the function first, followed by the
# arguments in the order in which they are called in `FUN`
mapply( FUN = function(x, y) x * y
, 1:10
, rep( x = 2:4, each = 10)
)
# [1] 2 4 6 8 10 12 14 16 18 20 3 6 9 12 15 18 21 24 27 30 4 8 12 16 20
# [26] 24 28 32 36 40
Run Code Online (Sandbox Code Playgroud)
首先,lapply()如果你的函数是矢量化的,你可以这样做.在这种情况下,它是:
x <- 1:10
unlist(lapply(2:4, function(y) x*y))
# OR
unlist(lapply(2:4, function(x=x,y) x*y))
Run Code Online (Sandbox Code Playgroud)
其次,如果需要对两个向量的每个组合应用函数,请使用outer():
xf <- 1:10
yf <- 2:4
c(xf %o% yf)
# OR spelled out for any function:
c(outer(xf,yf,FUN = `*`))
Run Code Online (Sandbox Code Playgroud)
如果使用mapply,则可以使用参数MoreArgs来避免必须使用rep构造参数:
xf <- 1:10
yf <- 2:4
mapply(function(x,y) x*y,
y = yf,
MoreArgs = list(x = xf))
Run Code Online (Sandbox Code Playgroud)
这与lapply()我上面显示的构造完全等效.生成的矩阵也可以使用SIMPLIFY = FALSE和转换为矢量unlist():
unlist(mapply(function(x,y) x*y,
y = yf,
MoreArgs = list(x = xf),
SIMPLIFY = FALSE))
Run Code Online (Sandbox Code Playgroud)
哪种解决方案最方便,取决于您的实际用例.时间方面它们都是可比较的,在最近的R版本中可能outer()比其他解决方案稍慢.
为了说明结果如何根据对象的大小和顺序而大不相同,我包括以下基准测试结果(下面的代码和输出).这表明:
outer() 不一定是最快的解决方案,尽管它通常是最快的解决方案之一.mapply()会增加很多开销,即使双重sapply()调用也会更快.代码:警告:这将运行一段时间
fx <- sample(1e4)
fy <- sample(1e3)
library(microbenchmark)
microbenchmark(sapply = as.vector(sapply(fx, function(x, y) x * y, fy)),
mapply = mapply( FUN = function(x, y) x * y, fx, rep( fy, each = 1e4)),
sapply2 = as.vector(sapply(fx, function(y) sapply(fy, function(x) x * y))),
outer = c(outer(fx, fy, function(x, y) x * y)),
mapply2 = mapply(function(x,y) x*y, x=fx, MoreArgs = list(y = fy)),
mapply3 = mapply(function(x,y) x*y, y=fy, MoreArgs = list(x = fx)),
times = 15)
Run Code Online (Sandbox Code Playgroud)
我机器上的输出:
Unit: milliseconds
expr min lq mean median uq max neval cld
sapply 89.52318 92.98653 344.1538 117.11280 239.64887 1485.3178 15 a
mapply 20471.02137 22925.42757 24478.5985 24650.29055 25627.31232 28840.3494 15 c
sapply2 7472.02251 8268.04696 9519.8016 8707.19193 9528.46181 14182.7537 15 b
outer 77.62331 85.94651 189.5107 91.83722 182.08506 1119.6620 15 a
mapply2 77.76871 79.71924 143.9484 81.24168 84.53247 971.1792 15 a
mapply3 65.21709 71.85662 107.9586 73.80779 124.21141 242.0760 15 a
Run Code Online (Sandbox Code Playgroud)
试试outer:
c(outer(1:10, 2:4, Vectorize(function(x, y) x*y)))
## [1] 2 4 6 8 10 12 14 16 18 20 3 6 9 12 15 18 21 24 27 30 4 8 12 16 20
## [26] 24 28 32 36 40
Run Code Online (Sandbox Code Playgroud)
如果函数已经被矢量化,就像在这里一样,那么我们可以省略Vectorize:
c(outer(1:10, 2:4, function(x, y) x * y))
## [1] 2 4 6 8 10 12 14 16 18 20 3 6 9 12 15 18 21 24 27 30 4 8 12 16 20
## [26] 24 28 32 36 40
Run Code Online (Sandbox Code Playgroud)
实际上,在这种特殊情况下,显示的匿名函数是默认函数,因此这将起作用:
c(outer(1:10, 2:4))
## [1] 2 4 6 8 10 12 14 16 18 20 3 6 9 12 15 18 21 24 27 30 4 8 12 16 20
## [26] 24 28 32 36 40
Run Code Online (Sandbox Code Playgroud)
在这种特殊情况下我们也可以使用:
c(1:10 %o% 2:4)
## [1] 2 4 6 8 10 12 14 16 18 20 3 6 9 12 15 18 21 24 27 30 4 8 12 16 20
## [26] 24 28 32 36 40
Run Code Online (Sandbox Code Playgroud)
如果您的起点是问题中X显示的列表,那么:
c(outer(X[[1]], X[[2]], Vectorize(function(x, y) x * y)))
## [1] 2 4 6 8 10 12 14 16 18 20 3 6 9 12 15 18 21 24 27 30 4 8 12 16 20
## [26] 24 28 32 36 40
Run Code Online (Sandbox Code Playgroud)
要么
c(do.call("outer", c(unname(X), Vectorize(function(x, y) x*y))))
## [1] 2 4 6 8 10 12 14 16 18 20 3 6 9 12 15 18 21 24 27 30 4 8 12 16 20
## [26] 24 28 32 36 40
Run Code Online (Sandbox Code Playgroud)
如果适用,前面部分适用于缩短它的地方.