left_join R dataframes, merging two columns with NAs

问题

My problem is the following: Lets say I have an existing dataframe with the following columns: UID, foo, result. Result is already partially filled. A second model now predicts additional rows, generating a second dataframe containing a UID and a result column: (Code to reproduce at bottom)

## df_main
##    UID   foo result
##  <dbl> <chr>  <chr>
## 1     1   moo    Cow
## 2     2   rum   <NA>
## 3     3  oink   <NA>
## 4     4  woof    Dog
## 5     5  hiss   <NA>

## new_prediction
##    UID result
##  <dbl>  <chr>
## 1     3    Pig
## 2     5  Snake

I now want to left_join the new results by UID to get the following result column:

## Cow
## <NA>
## Pig
## Dog
## Snake

But I can't get that to work, since left_join(df_main, new_prediction, by="UID") creates result.x and result.y. Is there any way to do this with dplyr, or alternatively, a good second step to join the columns? I looked at various functions, but finally resolved to loop over all rows manually. I am pretty certain that there is a more "R" way to do that?

Code for dataframes:

df_main <- tibble(UID = c(1,2,3,4,5), foo=c("moo", "rum", "oink", "woof", "hiss"), result=c("Cow", NA, NA, "Dog", NA))
new_prediction <- tibble(UID = c(3,5), result = c("Pig", "Snake"))

回答1:

coalesce is your second step.

left_join(df_main, new_prediction, by="UID") %>%
  mutate(result = coalesce(result.x, result.y)) %>%
  select(-result.x, -result.y)
# # A tibble: 5 x 3
#     UID   foo result
#   <dbl> <chr>  <chr>
# 1     1   moo    Cow
# 2     2   rum   <NA>
# 3     3  oink    Pig
# 4     4  woof    Dog
# 5     5  hiss  Snake

coalesce will accept as many columns as you give it. Earlier columns have precedence in case there are multiple non-missing values.

回答2:

Adding to Gregor's answer of using coalesce, you could also "manually" join the columns with ifelse.

left_join(df_main, new_prediction, by = "UID") %>%
  mutate(result = ifelse(is.na(result.x),result.y, result.x)) %>%
  select(-c(result.x, result.y))
# A tibble: 5 x 3
# UID foo   result
# <dbl> <chr> <chr> 
# 1  1.00 moo   Cow   
# 2  2.00 rum   <NA>  
# 3  3.00 oink  Pig   
# 4  4.00 woof  Dog   
# 5  5.00 hiss  Snake

来源：https://stackoverflow.com/questions/48284524/left-join-r-dataframes-merging-two-columns-with-nas

标签

join

dplyr

left-join