跨列填充data.frame中的缺失值

Question

Normally people want to fill missing values up or down a column. I would like to fill across, left or right.

# all character data.frame, except 1st column
df <- data.frame(a = 1:4, 
                 b = c('row1', 'row2', 'row3', 'row4'),
                 c = paste(9:12),
                 d = paste(13:16))
# remove a few values
df[2,2] <- df[3,3] <- df[4,2] <- NA

> df
  a    b    c  d
1 1 row1    9 13
2 2 <NA>   10 14
3 3 row3 <NA> 15
4 4 <NA>   12 16

# fill down. This is straighforward and works as expected.
df%>%fill(names(.), .direction='down')

How do I fill across???

# this doesn't work
df%>%fill(names(.), direction='right')

Lets assume we coerce to character if the fill value does not match the type of the missing value. But only for the columns that require coercion. So column a should stay numeric

Answer 1

This would be the equivalent of the right variant:

library(tidyverse)

df %>% 
  rowid_to_column %>% 
  gather(key, val, -rowid) %>% 
  arrange(rowid) %>%
  fill(val) %>% 
  spread(key, val) %>% select(-rowid)

Basically you can turn the data into long format and then use fill .

Direction down is then equivalent to right and direction up equivalent to left if you use the code above.

Output:

  a    b    c  d
1 1 row1    9 13
2 2    2   10 14
3 3 row3 row3 15
4 4    4   12 16

Here the attributes are dropped, and you'd need to re-establish the type of column you want.

Answer 2

This solution transposes the data frame, fills down, then transposes back again. The transposition converts the data frame to a matrix, so requires that it is converted back.

df <- data.frame(a = 1:4, 
                 b = c('row1', 'row2', 'row3', 'row4'),
                 c = paste(9:12),
                 d = paste(13:16))
# remove a few values
df[2,2] <- df[3,3] <- df[4,2] <- NA

library(tidyverse)

t(df) %>% 
  as.data.frame %>% 
  fill(names(.),.direction = "down") %>% 
  t %>% 
  as.data.frame
#>    a    b    c  d
#> V1 1 row1    9 13
#> V2 2    2   10 14
#> V3 3 row3 row3 15
#> V4 4    4   12 16

^{Created on 2019-02-08 by the reprex package (v0.2.1.9000)}