r - 条件(不等式)加入 data.table

Question

我只是想弄清楚如何对两个 data.tables 进行条件连接。





我写了一个 sqldf有条件的加入给我开始或结束时间在另一个开始/结束时间之内的电路。

sqldf("select dt2.start, dt2.finish, dt2.counts, dt1.id, dt1.circuit 
      from dt2 
        left join dt1 on (
          (dt2.start  >= dt1.start and dt2.start  < dt1.finish) or 
          (dt2.finish >= dt1.start and dt2.finish < dt1.finish)
        )")

这给了我正确的结果，但是对于我的大型数据集来说太慢了。

什么是data.table在没有矢量扫描的情况下如何做到这一点？

这是我的数据:

dt1 <- data.table(structure(list(circuit = structure(c(2L, 1L, 2L, 1L, 2L, 3L, 
1L, 1L, 2L), .Label = c("a", "b", "c"), class = "factor"), start = structure(c(1393621200, 
1393627920, 1393628400, 1393631520, 1393650300, 1393646400, 1393656000, 
1393668000, 1393666200), class = c("POSIXct", "POSIXt"), tzone = ""), 
    end = structure(c(1393626600, 1393631519, 1393639200, 1393632000, 
    1393660500, 1393673400, 1393667999, 1393671600, 1393677000
    ), class = c("POSIXct", "POSIXt"), tzone = ""), id = structure(1:9, .Label = c("1001", 
    "1002", "1003", "1004", "1005", "1006", "1007", "1008", "1009"
    ), class = "factor")), .Names = c("circuit", "start", "end", 
"id"), class = "data.frame", row.names = c(NA, -9L)))


dt2 <- data.table(structure(list(start = structure(c(1393621200, 1393624800, 1393626600, 
1393627919, 1393628399, 1393632000, 1393639200, 1393646399, 1393650299, 
1393655999, 1393660500, 1393666199, 1393671600, 1393673400), class = c("POSIXct", 
"POSIXt"), tzone = ""), end = structure(c(1393624799, 1393626600, 
1393627919, 1393628399, 1393632000, 1393639200, 1393646399, 1393650299, 
1393655999, 1393660500, 1393666199, 1393671600, 1393673400, 1393677000
), class = c("POSIXct", "POSIXt"), tzone = ""), seconds = c(3599L, 
1800L, 1319L, 480L, 3601L, 7200L, 7199L, 3900L, 5700L, 4501L, 
5699L, 5401L, 1800L, 3600L), counts = c(1L, 1L, 0L, 1L, 2L, 1L, 
0L, 1L, 2L, 3L, 2L, 3L, 2L, 1L)), .Names = c("start", "end", 
"seconds", "counts"), row.names = c(1L, 3L, 4L, 5L, 6L, 7L, 8L, 
9L, 10L, 11L, 12L, 13L, 14L, 15L), class = "data.frame"))

Answer 1

使用非 equi 连接:

ans = dt1[dt2, on=.(start <= end, end > start), 
           .(i.start, i.end, counts, id, circuit, cndn = i.start < x.start & i.end >= x.end), 
           allow.cartesian=TRUE
        ][!cndn %in% TRUE]

条件start <= end, end >= start (注意两种情况下的>=)将检查两个间隔是否以任何方式重叠。一侧的开区间由 end > start 完成部分( > 而不是 >= )。但它仍然会选择类型的间隔:

         dt1: start=================end
   dt2: start--------------------------------end ## start < start, end > end

和

         dt1: start=================end
                dt2: start----------end          ## end == end

cndn列是检查和删除这些情况。希望这些情况不会很多，这样我们就不会不必要地实现不需要的行。

PS:这种情况下的解决方案并不像我想的那么简单，那是因为该解决方案需要 OR手术。可以进行两个条件连接，然后将它们绑定(bind)在一起。

也许在某些时候，我们必须考虑以更直接的方式将连接扩展到这些类型的操作的可行性。