data.table中的重塑难题

Question

又一个重塑问题了 data.table

set.seed(1234)
DT <- data.table(x=rep(c(1,2,3),each=4), y=c("A","B"), v=sample(1:100,12))
#    x y  v
# 1: 1 A 12
# 2: 1 B 62
...
#11: 3 A 63
#12: 3 B 49

我想这样做的累加值x和v通过y,但结果呈现为:行数始终保持不变,而当y==A在SUM.*.A递增,同一时候y==B.(像往常一样y可能有很多因素,本例中为2)

#     SUM.x.A SUM.x.B  SUM.v.A SUM.v.B
# 1:        1      NA       12      NA
# 2:        1       1       12      62
...
#11:       12       9       318     289
#12:       12      12       318     338

编辑:这是我的糟糕解决方案显然过于复杂

#first step is to create cumsum columns
colNames <- c("x","v"); newColNames <- paste0("SUM.",colNames)
DT[, newColNames:=lapply(.SD,cumsum) ,by=y, .SDcols=colNames, with=F];
#now we need to reshape each SUM.* to SUM.*.{yvalue}
DT[,N:=.I]; setattr(DT,"sorted","N")

g <- function(DT,SD){
  cols <- c('N',grep('SUM',colnames(SD), value=T));
  Yval <- unique(SD[,y]);
  merge(DT, SD[,cols, with=F], suffixe=c('',paste0('.',Yval)), all.x=T);    
}

DT <- Reduce(f=g,init=DT,x=split(DT,DT$y));

locf = function(x) {
  ind = which(!is.na(x))    
  if(is.na(x[1])) ind = c(1,ind)
  rep(x[ind], times = diff( c(ind, length(x) + 1) )) 
}

newColNames <- grep('SUM',colnames(DT),value=T);
DT <- DT[, (newColNames):=lapply(.SD, locf), .SDcols=newColNames]

Answer 1

试试这个:

cumsum0 <- function(x) { x <- cumsum(x); ifelse(x == 0, NA, x) }
DT2 <- DT[, {SUM.<-y; lapply(data.table(model.matrix(~ SUM.:x + SUM.:v + 0)), cumsum0)}]
setnames(DT2, sub("(.):(.)", "\\2.\\1", names(DT2)))

简化:

1)如果使用0代替NA是正确的,那么可以通过省略第一行来简化它,该第一行在下一行中定义cumsum0和替换. cumsum0cumsum

2)第二行的结果有以下名称:

> names(DT2)
[1] "SUM.A:x" "SUM.B:x" "SUM.A:v" "SUM.B:v"

因此,如果这足够,则可以删除最后一行,因为它的唯一目的是使名称与问题中的名称完全相同.

结果(没有简化)是:

> DT2
    SUM.x.A SUM.x.B SUM.v.A SUM.v.B
 1:       1      NA      12      NA
 2:       1       1      12      62
 3:       2       1      72      62
 4:       2       2      72     123
 5:       4       2     155     123
 6:       4       4     155     220
 7:       6       4     156     220
 8:       6       6     156     242
 9:       9       6     255     242
10:       9       9     255     289
11:      12       9     318     289
12:      12      12     318     338