Join or Merge two data-frames based on multiple columns criteria - r

Question

I have two datasets with the following data.
maindata = data.frame(eventid=c(1:10),
district=c(rep("lucknow",2),rep("allahabad",1), rep("kanpur", 2)),
date = c(rep("2018-01-01", 2), rep("2018-01-02", 1), rep("2018-01-03", 2)))
weather = data.frame(district=c(rep("lucknow", 4), rep("allahabad", 3), rep("kanpur", 3)),
date = c(rep("2017-01-01", 4), rep("2017-01-02", 3), rep("2017-01-03", 3)),
temperature=c(rep("19.3",2),rep("22.1",1), rep("24.1", 2)))
Few considerations:
"date" in each data frame is different, its ok to be like that. MM-DD are sufficient
Both datasets have different length - df1 is my main dataset where "temp" should be added
The merging must happen over "district" and "date"
maindata has district column in lowercase
What i Tried: (doing some silly conversions.. will fix them)
weather$District<-as.factor(tolower(weather$District))
weather$Date<-as.Date(as.character(weather$Date),format="%m/%d/%Y")
maindata$md<-strftime(data$createDate, "%m-%d")
weather$mdr<-strftime(weather$Date, "%m-%d")
maindata<-left_join(maindata, weather, by = c("md" = "mdr", "district" = "District"))
The final expected answer would be something like below in maindata
eventid district date temperature
1 lucknow 2018-01-01 19.3
2 lucknow 2018-01-01 19.3
3 allahabad 2018-01-03 24.1
4 kanpur 2018-01-03 NA
5 kanpur 2018-01-02 22.1
6 lucknow 2018-01-01 19.3
7 lucknow 2018-01-01 19.3
8 allahabad 2018-01-03 24.1
9 kanpur 2018-01-03 NA
10 kanpur 2018-01-02 22.1
Can anybody please help !!!

Answer 1

I don't understand your logic rules for merging; specifically I don't see how date comes in.
It is entirely possible to reproduce your expected output without considering date at all, by simply matching df1$district with df2$dist:
library(tidyverse);
left_join(df1, df2, by = c("district" = "dist")) %>%
distinct() %>%
select(-date.y)
# eventid date.x district temp
#1 1 2017-01-01 dist-1 19.3
#2 2 2017-01-01 dist-1 19.3
#3 3 2017-01-01 dist-1 19.3
#4 4 2017-01-01 dist-1 19.3
#5 5 2017-01-02 dist-2 22.1
#6 6 2017-01-02 dist-2 22.1
#7 7 2017-01-02 dist-2 22.1
#8 8 2017-01-03 dist-3 24.10
#9 9 2017-01-03 dist-3 24.10
#10 10 2017-01-03 dist-3 24.10
Could you provide sample data that is more representative of what you're trying to do, and where the role/importance of merging on date becomes clear?

Answer 2

A quick note - You should really post your trials to the solution before asking for the help in SO.
To the answer -
What you should be using is the merge function available by default in R.
After reproducing the data frames that you have provided - try the below chunk of code
#Since dates doesn't matter, df2 could be changed to a new df with only temp
df3 <- df2[,c("dist","temp")]
df3 <- unique(df3)
df4 <- merge(df1,df3,by.x = "district",by.y = "dist",all.x = T)
The deduplication has been done to avoid creation of numerous rows for each combination of dates in df1 and df2.
all.x = T ensures that you're getting a left-join (Where all rows of the df1 are present in your final output)

Answer 3

Perhaps something like this (with the updated data)
library(tidyverse)
df1 %>%
mutate(date = as.POSIXct(date),
date1 = format(date, "%d/%m")) %>%
left_join(df2 %>%
mutate(date = as.POSIXct(date),
date1 = format(date, "%d/%m")), by = c("date1" = "date1", "district" = "dist")) %>%
select(-date1, - date.y) %>%
rename(date = date.x) %>%
filter(!duplicated(eventid))
#output
eventid date district temp
1 1 2017-01-01 dist-1 19.3
2 2 2017-01-01 dist-1 19.3
3 3 2017-01-01 dist-1 19.3
4 4 2017-01-01 dist-1 19.3
5 5 2017-01-02 dist-2 <NA>
6 6 2017-01-02 dist-2 <NA>
7 7 2017-01-02 dist-2 <NA>
8 8 2017-01-03 dist-3 24.10
9 9 2017-01-03 dist-3 24.10
10 10 2017-01-03 dist-3 24.10
Convert date in both data frames to POSIXct, make a %d/%m column and join by it and district, and then clean up

Answer 4

Maybe you want this.
df2[, 2] <- as.numeric(as.character(df2[, 2]))
m1 <- merge(df1, df2, by.x = "district", by.y = "dist", all.x = TRUE)[-5]
names(m1)[3] <- "date"
m1 <- unique(m1[, c(2, 3, 1, 4)])
rownames(m1) <- NULL
> m1
eventid date district temp
1 1 2017-01-01 dist-1 19.3
2 2 2017-01-01 dist-1 19.3
3 3 2017-01-01 dist-1 19.3
4 4 2017-01-01 dist-1 19.3
5 5 2017-01-02 dist-2 22.1
6 6 2017-01-02 dist-2 22.1
7 7 2017-01-02 dist-2 22.1
8 8 2017-01-03 dist-3 24.1
9 9 2017-01-03 dist-3 24.1
10 10 2017-01-03 dist-3 24.1