Question

I have found myself doing a “conditional left join” several times in R. To illustrate with an example; if you have two data frames such as:

> df
    a b
  1 1 0
  2 2 0

> other.df
    a b
  1 2 3

The goal is to end up with this data frame:

> final.df
    a b
  1 1 0
  2 2 3

The code I’ve been written so far:

c <- merge(df, other.df, by=c("a"), all.x = TRUE)
c[is.na(c$b.y),]$b.y <- 0
d<-subset(c, select=c("a","b.y"))
colnames(d)[2]<-b

to finally arrive with the result I wanted.

Doing this in effectively four lines makes the code very opaque.
Is there any better, less cumbersome way to do this?

Answer 1

Here are two ways. In both cases the first line does a left merge returning the required columns. In the case of merge we then have to set the names. The final line in both lines replaces NAs with 0.

merge

res1 <- merge(df, other.df, by = "a", all.x = TRUE)[-2]
names(res1) <- names(df)
res1[is.na(res1)] <- 0

sqldf

library(sqldf)
res2 <- sqldf("select a, o.b from df left join 'other.df' o using(a)")
res2[is.na(res2)] <- 0