时间戳

首页 > 解决方案 > 在数据表列表上应用一个函数，该函数需要引用上一行

r - 在数据表列表上应用一个函数，该函数需要引用上一行

问题描述

我想对数据表的所有行应用一个函数，对于列表中的所有数据表，但是该函数必须引用计算中的前一行（除了第 1 行，它是独立的）

当不需要参考上一行时，我可以这样做；

require(data.table)

# dummy list data
l <- list(data.table(col1=c(2,3,4,2,1), col2=c(1,2,3,4,3), col3=c(5,4,3,4,5), col4=c(1,1,1,1,2)), data.table(col1=c(3,4,3,2,3), col2=c(1,3,4,2,2), col3=c(5,4,3,2,3), col4=c(5,5,5,5,5)))

# apply the function to the data table by row, and apply this function to all tables in the list
lapply(l, function(b) b[ , value := mapply(function(w,x,y,z) w + x + y * z, col1, col2, col3, col4)])

但是我该如何value考虑value上一行？

# this wont work b'cos value hasn't been created yet
lapply(l, function(b) b[ , value := mapply(function(w,x,y,z) w + x + y * z, col1, col2, col3, col4)])

 Error in mapply(function(w, x, y, z, v) (w + x + y * z)/shift(v, 1), col1,  : 
  object 'value' not found 

# so make 'value' for row 1 only and try again
lapply(l, function(b) b[ , value := 0])
lapply(l, function(b) b[1 , value := col1 + col2 + col3 - col4])

# using shift inside mapply
lapply(l, function(b) b[ , value := mapply(function(w,x,y,z,v) (w + x + y * z) / shift(v,1), col1, col2, col3, col4, value)])

将value列全部转换为 NA

标签： rdata.tableapply

解决方案

updated to reflect other comments

for this case couldn't you just avoid apply functions altogether and do something like:

names(l) <- c("list1","list2")
l <- rbindlist(l,idcol=TRUE)
l[,value := col1 + col2 + col3 * col4,by=.id]
l[,value2 := value / shift(value, fill = 1),by=.id]

which returns 

      .id col1 col2 col3 col4 value     value2
 1: list1    2    1    5    1     8  8.0000000
 2: list1    3    2    4    1     9  1.1250000
 3: list1    4    3    3    1    10  1.1111111
 4: list1    2    4    4    1    10  1.0000000
 5: list1    1    3    5    2    14  1.4000000
 6: list2    3    1    5    5    29 29.0000000
 7: list2    4    3    4    5    27  0.9310345
 8: list2    3    4    3    5    22  0.8148148
 9: list2    2    2    2    5    14  0.6363636
10: list2    3    2    3    5    20  1.4285714

the important part here is the fill argument of the shift() function, described here

推荐阅读