data.table：第一次出现的行(data.table: field in row with first ocurrence)

我想探索一个data.table分组的优雅（单行）解决方案

我有一个data.table如下：

library(data.table) library(lubridate) dt.master <- data.table(user = c(1000, 1002, 2008, 3005, 1000, 1002, 1002), target = c(50000, 50004, 50501, 50001, 50000, 50000, 50004), channel = c("A", "B", "C", "A", "B", "A", "C"), date = c(dmy("10/02/2018"), dmy("11/04/2018"), dmy("14/03/2018"), dmy("02/03/2018"), dmy("05/01/2018"), dmy("08/05/2018"), dmy("05/03/2018")))

那是：

user target channel date 1: 1000 50000 A 2018-02-10 2: 1002 50004 B 2018-04-11 3: 2008 50501 C 2018-03-14 4: 3005 50001 A 2018-03-02 5: 1000 50000 B 2018-01-05 6: 1002 50000 A 2018-05-08 7: 1002 50004 C 2018-03-05

我想知道，对于每个(user, target)组的第一次出现的频道，并将其添加到dt.master。这是：

user target channel date first_channel 1: 1000 50000 A 2018-02-10 B 2: 1000 50000 B 2018-01-05 B 3: 1002 50000 A 2018-05-08 A 4: 1002 50004 B 2018-04-11 C 5: 1002 50004 C 2018-03-05 C 6: 2008 50501 C 2018-03-14 C 7: 3005 50001 A 2018-03-02 A

目前，我正在分两步进行：

首先，我提取第一次发生的行

dt.result <- dt.master[dt.master[, .(first_interest = .I[which.min(date)]), by = c("user", "target")]$first_interest,]

之后，我将它与dt.master合并

setnames(dt.result, "channel", "first_channel") dt.master <- merge(dt.master, dt.result[, .(user, target, first_channel)], by.x = c("user", "target"), by.y = c("user", "target"), all.x = T, all.y = F)

有没有办法做到没有合并？ 我相信必须有修改第一行的解决方案，但我找不到它。

非常感谢！

I would like to explore an elegant (one-liner) solution for a data.table grouping

I have a data.table as follows:

library(data.table) library(lubridate) dt.master <- data.table(user = c(1000, 1002, 2008, 3005, 1000, 1002, 1002), target = c(50000, 50004, 50501, 50001, 50000, 50000, 50004), channel = c("A", "B", "C", "A", "B", "A", "C"), date = c(dmy("10/02/2018"), dmy("11/04/2018"), dmy("14/03/2018"), dmy("02/03/2018"), dmy("05/01/2018"), dmy("08/05/2018"), dmy("05/03/2018")))

That is:

user target channel date 1: 1000 50000 A 2018-02-10 2: 1002 50004 B 2018-04-11 3: 2008 50501 C 2018-03-14 4: 3005 50001 A 2018-03-02 5: 1000 50000 B 2018-01-05 6: 1002 50000 A 2018-05-08 7: 1002 50004 C 2018-03-05

I would like to know, for each group of (user, target), the channel of the first ocurrence, and add it to the dt.master. This is:

user target channel date first_channel 1: 1000 50000 A 2018-02-10 B 2: 1000 50000 B 2018-01-05 B 3: 1002 50000 A 2018-05-08 A 4: 1002 50004 B 2018-04-11 C 5: 1002 50004 C 2018-03-05 C 6: 2008 50501 C 2018-03-14 C 7: 3005 50001 A 2018-03-02 A

Currently, I am doing it in two steps:

First, I extract the rows of the first ocurrence

dt.result <- dt.master[dt.master[, .(first_interest = .I[which.min(date)]), by = c("user", "target")]$first_interest,]

And, afterwards, I merge it with dt.master

setnames(dt.result, "channel", "first_channel") dt.master <- merge(dt.master, dt.result[, .(user, target, first_channel)], by.x = c("user", "target"), by.y = c("user", "target"), all.x = T, all.y = F)

Is there a way to do it with no merges? I believe there must be a solution modifying the first line, but I cannot find it.

Thanks a lot!