R：有条件地将数据从一个数据帧提取到另一个数据帧

问题描述

我有两个数据框，我想有条件地从一个数据框的一列中提取数据，然后将其放入另一个数据框的新列中。

数据帧1看起来像这样：

df1 <- data.frame(date.start = c("2019-06-10 11:52:00","2019-06-11 11:52:00","2019-06-12 11:51:00"),date.end =
  c("2019-06-10 11:53:00","2019-06-11 11:53:00","2019-06-12 11:53:00"))

数据框2如下所示：

df2 <- data.frame(date.start = c("2019-06-11 11:50:00","2019-06-10 11:51:00","2019-06-12 11:50:00"),date.end =
  c("2019-06-11 11:54:00","2019-06-11 08:59:00","2019-06-12 11:57:00"),day = c(1,15,64))

如果df.1的date.start和date.end落在date.start任何行的date.end或df2内，我要提取变量{ day中的{1}}，并将其放入df2的匹配行中。

预期结果如下：

df1

我目前有以下循环可以正常工作，但是当我在大型数据帧（行= 1135133）上运行它时，它的运行速度非常慢，我想知道是否有更快的方法可以实现这一点。

expected.out <- data.frame(date.start = c("2019-06-10 11:52:00",date.end = c("2019-06-10 11:53:00","2019-06-12 11:53:00"),day = c(15,1,64))

解决方法

使用library(fuzzyjoin)

library(tidyverse)
library(lubridate)
library(fuzzyjoin)

df1 <- data.frame(
  date.start = c("2019-06-10 11:52:00","2019-06-11 11:52:00","2019-06-12 11:51:00"),date.end = c("2019-06-10 11:53:00","2019-06-11 11:53:00","2019-06-12 11:53:00"),stringsAsFactors = F)

df2 <- data.frame(date.start = c("2019-06-11 11:50:00","2019-06-10 11:51:00","2019-06-12 11:50:00"),date.end = c("2019-06-11 11:54:00","2019-06-11 08:59:00","2019-06-12 11:57:00"),day = c(1,15,64),stringsAsFactors = F)

df1 <- df1 %>% 
  mutate(across(where(is.character),ymd_hms)) %>% 
  as_tibble()

df2 <- df2 %>% 
  mutate(across(where(is.character),ymd_hms)) %>% 
  as_tibble()


fuzzy_left_join(df1,df2,by = c("date.start","date.end"),match_fun = list(`>=`,`<=`))
# A tibble: 3 x 5
  date.start.x        date.end.x          date.start.y        date.end.y            day
  <dttm>              <dttm>              <dttm>              <dttm>              <dbl>
1 2019-06-10 11:52:00 2019-06-10 11:53:00 2019-06-10 11:51:00 2019-06-11 08:59:00    15
2 2019-06-11 11:52:00 2019-06-11 11:53:00 2019-06-11 11:50:00 2019-06-11 11:54:00     1
3 2019-06-12 11:51:00 2019-06-12 11:53:00 2019-06-12 11:50:00 2019-06-12 11:57:00    64

^{由reprex package（v0.3.0）于2020-09-23创建}

不确定该方法是否快速

您可以在match内使用sapply来获取df2的第一行，其中日期在给定时间范围内。

df1[] <- lapply(df1,as.POSIXct) #Convert character to POSIXct
df2[1:2] <- lapply(df2[1:2],as.POSIXct)

df1$day <- df2$day[sapply(asplit(df1,1),function(x) {match(TRUE,x[1] >= df2[,1] & x[2] <= df2[,2])})]
df1
#           date.start            date.end day
#1 2019-06-10 11:52:00 2019-06-10 11:53:00  15
#2 2019-06-11 11:52:00 2019-06-11 11:53:00   1
#3 2019-06-12 11:51:00 2019-06-12 11:53:00  64

将between中的data.table与outer一起使用。 which.max扫描匹配矩阵中的TRUE值。

library(data.table)
FUN <- Vectorize(function(x,y) all(between(unlist(df1[x,]),df2[y,1],2])))
res <- transform(df1,day=df2[apply(outer(1:3,1:3,FUN),1,which.max),3])
res
#            date.start            date.end day
# 1 2019-06-10 11:52:00 2019-06-10 11:53:00  15
# 2 2019-06-11 11:52:00 2019-06-11 11:53:00   1
# 3 2019-06-12 11:51:00 2019-06-12 11:53:00  64

您可能需要预先转换为POSIXct格式以应用解决方案。

df1[1:2] <- lapply(df1[1:2],as.POSIXct)
df2[1:2] <- lapply(df2[1:2],as.POSIXct)

数据：

df1 <- structure(list(date.start = structure(c(1560160320,1560246720,1560333060),class = c("POSIXct","POSIXt"),tzone = ""),date.end = structure(c(1560160380,1560246780,1560333180),tzone = "")),row.names = c(NA,-3L),class = "data.frame")

df2 <- structure(list(date.start = structure(c(1560246600,1560160260,1560333000),date.end = structure(c(1560246840,1560236340,1560333420),64)),class = "data.frame")

loops r r vectorization