如何使用purrr匹配查找表中的记录？

Question

I have this dataset

library(dplyr)
data_frame(Q1= c('AL', NA, 'TX', 'FL'), Q2=c('MN', 'CO', NA, NA), value=c(10,24,12,54)) 
# A tibble: 4 x 3
     Q1    Q2 value
  <chr> <chr> <dbl>
1    AL    MN    10
2  <NA>    CO    24
3    TX  <NA>    12
4    FL  <NA>    54

And I am trying to use purrr to convert the values in Q1 and Q2 into full state names using a lookup table

lktState <- data_frame(abb=state.abb, name=state.name)

So far I've tried this but it doesn't work

data_frame(Q1= c('AL', NA, 'TX', 'FL'), Q2=c('MN', 'CO', NA, NA), value=c(10,24,12,54)) %>% 
  mutate_at(vars('Q1','Q2'), purrr::map(.x = ., lktState$name[match(.x, lktState$abb)]))

Error in match(.x, lktState$abb) : object '.x' not found

Answer 1

base R version (which can be vectorized but this illustrates the concept):

xdf <- data.frame(
  Q1= c('AL', NA, 'TX', 'FL'),
  Q2 = c('MN', 'CO', NA, NA),
  value = c(10, 24, 12, 54),
  stringsAsFactors=FALSE
) -> xdf

xdf
##     Q1   Q2 value
## 1   AL   MN    10
## 2 <NA>   CO    24
## 3   TX <NA>    12
## 4   FL <NA>    54
lktState <- setNames(state.name, state.abb)

xdf$Q1 <- lktState[xdf$Q1]
xdf$Q2 <- lktState[xdf$Q2]

xdf
##        Q1        Q2 value
## 1 Alabama Minnesota    10
## 2    <NA>  Colorado    24
## 3   Texas      <NA>    12
## 4 Florida      <NA>    54

"tidyverse"

library(dplyr)

xdf <- data_frame(
  Q1= c('AL', NA, 'TX', 'FL'),
  Q2 = c('MN', 'CO', NA, NA),
  value = c(10, 24, 12, 54)
) -> xdf

xdf
## # A tibble: 4 x 3
##      Q1    Q2 value
##   <chr> <chr> <dbl>
## 1    AL    MN    10
## 2  <NA>    CO    24
## 3    TX  <NA>    12
## 4    FL  <NA>    54
lktState <- setNames(state.name, state.abb)

mutate_at(xdf, .vars=vars(-value), .funs=funs(lktState[.]))
## # A tibble: 4 x 3
##        Q1        Q2 value
##     <chr>     <chr> <dbl>
## 1 Alabama Minnesota    10
## 2    <NA>  Colorado    24
## 3   Texas      <NA>    12
## 4 Florida      <NA>    54

There's no need to use "apply"-like idioms to do this basic lookup table assignment.

Answer 2

I agree with Sotos that a join is the natural way to do this. However, your purrr solution is definitely fixable.

You are missing three things,

1. For anything other than a simple single function, you need to use funs in mutate_at .
2. map functions use ~ notation for anonymous functions.
3. You don't want to return a list, but rather a character vector, so use _chr variant.

.

mutate_at(df,
          vars('Q1', 'Q2'), 
          funs(purrr::map_chr(.x = ., ~lktState$name[match(.x, lktState$abb)])))

Gives:

 # A tibble: 4 x 3 Q1 Q2 value <chr> <chr> <dbl> 1 Alabama Minnesota 10 2 <NA> Colorado 24 3 Texas <NA> 12 4 Florida <NA> 54 

Data

df <- data_frame(Q1= c('AL', NA, 'TX', 'FL'), Q2=c('MN', 'CO', NA, NA), value=c(10,24,12,54))