内部连接期间的向量减法

Question

我有以下数据集

    structure(list(Time = c(0L, 0L, 0L, 0L, 0L, 200L, 200L, 200L, 
200L, 200L, 400L, 400L, 400L, 400L, 400L), AgentID = c(1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), AC = c("c", 
"c", "c", "c", "c", "c", "c", "c", "c", "c", "c", "c", "c", "c", 
"c"), Layer = c(0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 
0L, 0L, 0L), Type = c("b", "b", "b", "b", "b", "b", "b", "b", 
"b", "b", "b", "b", "b", "b", "b"), Data = c(0, 0, 0, 0, 0, 0.117073864, 
0.13028602, 0.11111003, -0.11538354, 0.07852934, 0.24280901, 
0.24271743, 0.21535376, -0.2213944, 0.23355752), SimulationID = c(4L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L), discountFactor = c(0L, 
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), N = c(80L, 
80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 
80L)), row.names = c(NA, -15L), class = c("data.table", "data.frame"
), .internal.selfref = <pointer: 0x55843aed0a70>)

我想做的是用 on=.(Layer,Type,AC,AgentID,Time,SimulationID,N,discountFactor) 对行进行分组，然后执行 nodeVelocity:=Data-oldData，其中 oldData 是具有 oldTime=Time-200 的 Data 列.

我试过了

d[d[,.(Time=Time+200,Data), by=.(Layer,Type,AC,AgentID,SimulationID,N,discountFactor)], 
  `:=` (nodeVelocity=x.Data-i.Data, iData=i.Data)
  ,on=.(Layer,Type,AC,AgentID,Time = Time,SimulationID,N,discountFactor)]

然而，在上面的代码中，从它的iData列可以看出，减法是通过i.Data的最后一行（i.Data =: oldData）完成的。为什么是这样？我期待一个逐元素的向量减法，但我得到的是向量减去另一个向量的最后一行。

Answer 1

这里有一个选项：

DT[, c("ri", "T2") := .(rowid(rleid(Time, Layer, Type, AC, AgentID, SimulationID, N, discountFactor)),
    Time+200L)]
DT[DT, on=.(Time=T2, Layer, Type, AC, AgentID, SimulationID, N, discountFactor, ri),
    diff := x.Data - i.Data]

输出：

    Time AgentID AC Layer Type        Data SimulationID discountFactor  N ri  T2        diff
 1:    0       1  c     0    b  0.00000000            4              0 80  1 200          NA
 2:    0       1  c     0    b  0.00000000            4              0 80  2 200          NA
 3:    0       1  c     0    b  0.00000000            4              0 80  3 200          NA
 4:    0       1  c     0    b  0.00000000            4              0 80  4 200          NA
 5:    0       1  c     0    b  0.00000000            4              0 80  5 200          NA
 6:  200       1  c     0    b  0.11707386            4              0 80  1 400  0.11707386
 7:  200       1  c     0    b  0.13028602            4              0 80  2 400  0.13028602
 8:  200       1  c     0    b  0.11111003            4              0 80  3 400  0.11111003
 9:  200       1  c     0    b -0.11538354            4              0 80  4 400 -0.11538354
10:  200       1  c     0    b  0.07852934            4              0 80  5 400  0.07852934
11:  400       1  c     0    b  0.24280901            4              0 80  1 600  0.12573515
12:  400       1  c     0    b  0.24271743            4              0 80  2 600  0.11243141
13:  400       1  c     0    b  0.21535376            4              0 80  3 600  0.10424373
14:  400       1  c     0    b -0.22139440            4              0 80  4 600 -0.10601086
15:  400       1  c     0    b  0.23355752            4              0 80  5 600  0.15502818

在您的第二个问题上，这是因为联接中的每个 i 行有多个 x 行。遍历连接结果后，使用最后一行。查看最近的post：

内部连接期间的向量减法

Vector-wise substraction during inner join

r

data.table