如何将函数应用于具有purrr的成对的列？

Question

I have the following dataframe resulting from a join with dplyr:

data_frame(id=1:4, a.x = c(1, NA, 3, 4), a.y = c(1, 2, 3, 4), b.x = c(NA, NA, 3, NA), b.y = c(2, 2, NA, 4)) 
# A tibble: 4 x 5
     id   a.x   a.y   b.x   b.y
  <int> <dbl> <dbl> <dbl> <dbl>
1     1     1     1    NA     2
2     2    NA     2    NA     2
3     3     3     3     3    NA
4     4     4     4    NA     4

And I would like to replace all the NAs in the columns ending with .x with the value from the columns ending with .y . Eventually, I would like to achieve this:

# A tibble: 4 x 5
     id   a.x   a.y   b.x   b.y
  <int> <dbl> <dbl> <dbl> <dbl>
1     1     1     1     2     2
2     2     2     2     2     2
3     3     3     3     3    NA
4     4     4     4     4     4

I tried with purrr something like this:

data_frame(id=1:4, a.x = c(1, NA, 3, 4), a.y = c(1, 2, 3, 4), b.x = c(NA, NA, 3, NA), b.y = c(2, 2, NA, 4)) %>%
  map2_dfr(.x = ends_with('.y'), .y = ends_with('.x'), ~ case_when(is.na(.x) ~ .y,
                                                                   TRUE ~ .x))

Which is wrong. The documentation is a bit confusing to me, I think the issue here is that .x expects a vector, but how can I pass a list of columns then?

Answer 1

A tidyr solution. We can gather the columns, separate by . , arrange by columns, fill the value toward up, unite columns, and finally spread the data frame to the original structure.

library(tidyverse)

dat2 <- dat %>%
  gather(Column, Value, -id) %>%
  separate(Column, into = c("Col1", "Col2")) %>%
  arrange(id, Col1, Col2) %>%
  group_by(id, Col1) %>%
  fill(Value, .direction = "up") %>%
  unite(Column, Col1, Col2, sep = ".") %>%
  spread(Column, Value) %>%
  ungroup()
dat2
## A tibble: 4 x 5
#      id   a.x   a.y   b.x   b.y
# * <int> <dbl> <dbl> <dbl> <dbl>
# 1     1  1.00  1.00  2.00  2.00
# 2     2  2.00  2.00  2.00  2.00
# 3     3  3.00  3.00  3.00 NA   
# 4     4  4.00  4.00  4.00  4.00

Or if the order of the columns in the data frame is good, we can use the transpose function from the data.table package, but be careful that the column types may change after the process.

dat2 <- dat %>%
  data.table::transpose() %>%
  fill(everything(), .direction = 'up') %>%
  data.table::transpose() %>%
  setNames(names(dat))
dat2
#   id a.x a.y b.x b.y
# 1  1   1   1   2   2
# 2  2   2   2   2   2
# 3  3   3   3   3  NA
# 4  4   4   4   4   4 

Or a solution using purrr to create subset that with column names ends_with "x" and "y" first, and then replace the original columns ends with "x".

dat_x <- dat %>% select(ends_with("x"))
dat_y <- dat %>% select(ends_with("y"))

dat[, grepl("x$", names(dat))] <- map2(dat_x, dat_y, ~ifelse(is.na(.x), .y, .x)) 
dat
# # A tibble: 4 x 5
#      id   a.x   a.y   b.x   b.y
#   <int> <dbl> <dbl> <dbl> <dbl>
# 1     1  1.00  1.00  2.00  2.00
# 2     2  2.00  2.00  2.00  2.00
# 3     3  3.00  3.00  3.00 NA   
# 4     4  4.00  4.00  4.00  4.00

DATA

dat <- data_frame(id=1:4, a.x = c(1, NA, 3, 4), a.y = c(1, 2, 3, 4), b.x = c(NA, NA, 3, NA), b.y = c(2, 2, NA, 4))