2

我有两个数据表，让我们称它们weights为values.
该weights表有 5 列，如下所示：

first POSIXct
late POSIXct
nodeid integer
aggid integer
weight numeric

该values表有这些列

nodeid integer
Date POSIXct
hour integer
value decimal

这个想法是生成一个新表，它将根据权重将节点的加权平均值转换为聚合节点。但是，权重会随着时间而变化，需要根据第一个和最晚的日期进行匹配。执行此操作的 SQL 语法看起来像这样

select v.Date, v.hour, w.aggid, sum(v.value*w.weight) as aggvalue
from values v inner join weights w
on v.nodeid=w.nodeid and v.date between w.first and w.late
group by aggid, date, hour

between鉴于SQL 语法中的逻辑，我不确定从哪里开始。这在 data.table 语法中是否可行，或者我需要将weights表转换为每天都有一行而不是使用范围？

这是一些示例数据（对不起，它太长了）......

values<-data.table(nodeid = c(1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 
2L, 3L, 3L, 3L, 3L, 3L, 4L, 4L, 4L, 4L, 4L, 5L, 5L, 5L, 5L, 5L, 
6L, 6L, 6L, 6L, 6L), Date = c("2013-07-02", "2013-07-02", "2013-07-05", 
"2013-07-08", "2013-07-10", "2013-07-02", "2013-07-02", "2013-07-05", 
"2013-07-08", "2013-07-10", "2013-07-02", "2013-07-02", "2013-07-05", 
"2013-07-08", "2013-07-10", "2013-07-02", "2013-07-02", "2013-07-05", 
"2013-07-08", "2013-07-10", "2013-07-02", "2013-07-02", "2013-07-05", 
"2013-07-08", "2013-07-10", "2013-07-02", "2013-07-02", "2013-07-05", 
"2013-07-08", "2013-07-10"), hour = c(1L, 2L, 23L, 2L, 2L, 1L, 
2L, 23L, 2L, 2L, 1L, 2L, 23L, 2L, 2L, 1L, 2L, 23L, 2L, 2L, 1L, 
2L, 23L, 2L, 2L, 1L, 2L, 23L, 2L, 2L), value = c(8.234, 3.218, 
0.787, 8.689, 6.218, 6.89, 1.914, 2.459, 6.683, 8.122, 0.281, 
1.136, 1.993, 7.27, 9.582, 5.777, 1.375, 9.204, 7.862, 0.633, 
2.433, 1.842, 7.178, 10.692, 1.417, 1.259, 2.619, 0.031, 6.744, 
5.941))

weights<-data.table(first = c("2013-07-01", "2013-07-01", "2013-07-01", 
"2013-07-01", "2013-07-01", "2013-07-01", "2013-07-08", "2013-07-08", 
"2013-07-08", "2013-07-08", "2013-07-08", "2013-07-08"), late = c("2013-07-07", 
"2013-07-07", "2013-07-07", "2013-07-07", "2013-07-07", "2013-07-07", 
"2013-07-20", "2013-07-20", "2013-07-20", "2013-07-20", "2013-07-20", 
"2013-07-20"), nodeid = c(1L, 2L, 3L, 4L, 5L, 6L, 1L, 2L, 3L, 
4L, 5L, 6L), aggid = c(1L, 1L, 1L, 2L, 2L, 2L, 1L, 1L, 1L, 2L, 
2L, 2L), weight = c(0.5, 0.25, 0.25, 0.3, 0.5, 0.2, 0.6, 0.2, 
0.2, 0.4, 0.45, 0.15))

exresults<-data.table(aggid = c(1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L, 1L, 
2L), Date = c("2013-07-02", "2013-07-02", "2013-07-02", "2013-07-02", 
"2013-07-05", "2013-07-05", "2013-07-08", "2013-07-08", "2013-07-10", 
"2013-07-10"), hour = c(1L, 1L, 2L, 2L, 23L, 23L, 2L, 2L, 2L, 
2L), aggvalue = c(5.90975, 3.2014, 2.3715, 1.8573, 1.5065, 6.3564, 
8.004, 8.9678, 7.2716, 1.782))
4

1 回答 1

2

使用连接roll参数data.table：

setkey(values, nodeid, Date)
setkey(weights, nodeid, late)

weights[values, roll = -Inf][, list(aggvalue = sum(weight*value)),
                               by = list(aggid, Date = late, hour)]
#    aggid       Date hour aggvalue
# 1:     1 2013-07-02    1  5.90975
# 2:     1 2013-07-02    2  2.37150
# 3:     1 2013-07-05   23  1.50650
# 4:     1 2013-07-08    2  8.00400
# 5:     1 2013-07-10    2  7.27160
# 6:     2 2013-07-02    1  3.20140
# 7:     2 2013-07-02    2  1.85730
# 8:     2 2013-07-05   23  6.35640
# 9:     2 2013-07-08    2  8.96780
#10:     2 2013-07-10    2  1.78200

注意：如果不存在正确的范围，我会小心 - 我没有测试那个边缘情况。

于 2013-07-25T20:11:30.557 回答