Automatically coerce all column types of one data frame to the type of another prior to binding

Question

Let's say I have two data frames I want to bind:

ds_a <- data.frame(
  x = 1:6,
  y = 5,
  z = "4",
  l = 2,
  stringsAsFactors = FALSE
)

ds_b <- data.frame(
  x = as.factor(1:6),
  y = "5",
  p = 2,
  stringsAsFactors = FALSE
)

When I try to bind them I get the following error:

> bind_rows(ds_a, ds_b)
Error: Can't combine `..1$x` <integer> and `..2$x` <factor<4c79c>>.

Typically what I do to solve this is I convert all the columns in both data frames to a character, bind the two data frames, and then manually re-convert all the columns back to their original type.

Is there a way to simply coerce all the type collisions between ds_a and ds_b by automatically casting ds_b's columns to match ds_a (assuming they're named the same)?

More generally, I'd like a solution to automatically convert all the columns in ds_b to the type of ds_a wherever the column names match. And the solution should work if ds_b and ds_a don't share all the same columns (just filling with NA when columns don't exist in one, but do in another).

Here's the intended outcome:

ds_merged =read.table(text = 'x y z l p 
1 1 5 4 2 NA
2 2 5 4 2 NA
3 3 5 4 2 NA
4 4 5 4 2 NA
5 5 5 4 2 NA
6 6 5 4 2 NA
7 1 5 NA NA 2
8 2 5 NA NA 2
9 3 5 NA NA 2
10 4 5 NA NA 2
11 5 5 NA NA 2
12 6 5 NA NA 2', header = TRUE, row.names = NULL)

> ds_merged

   row.names x y  z  l  p
1          1 1 5  4  2 NA
2          2 2 5  4  2 NA
3          3 3 5  4  2 NA
4          4 4 5  4  2 NA
5          5 5 5  4  2 NA
6          6 6 5  4  2 NA
7          7 1 5 NA NA  2
8          8 2 5 NA NA  2
9          9 3 5 NA NA  2
10        10 4 5 NA NA  2
11        11 5 5 NA NA  2
12        12 6 5 NA NA  2

Answer 1

We could use type.convert()

Explanation: after comment of OP:

type_convert does not consider ds_a (you can check if you compare glimpse(ds_a) with glimpse of the resulting dataframe:

Note the columns of ds_a have the same classes as in result.

> # compare classes
> glimpse(ds_a)
Rows: 6
Columns: 4
$ x <int> 1, 2, 3, 4, 5, 6
$ y <dbl> 5, 5, 5, 5, 5, 5
$ z <chr> "4", "4", "4", "4", "4", "4"
$ l <dbl> 2, 2, 2, 2, 2, 2
> glimpse(ds_b)
Rows: 6
Columns: 3
$ x <fct> 1, 2, 3, 4, 5, 6
$ y <chr> "5", "5", "5", "5", "5", "5"
$ p <dbl> 2, 2, 2, 2, 2, 2
> glimpse(result)
Rows: 12
Columns: 5
$ x <int> 1, 2, 3, 4, 5, 6, 1, 2, 3, 4, 5, 6
$ y <dbl> 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5
$ z <chr> "4", "4", "4", "4", "4", "4", NA, NA, NA, NA, NA, NA
$ l <dbl> 2, 2, 2, 2, 2, 2, NA, NA, NA, NA, NA, NA
$ p <int> NA, NA, NA, NA, NA, NA, 2, 2, 2, 2, 2, 2

What type.convert does is:

1. to apply the best fitting class to the data of ds_b (notice the %>% is within bind_rows). So all of ds_b$x are integers therefore R converts class factor to class integer in ds_b$x.
2. All of ds_b$y are character class but integers in nature, therefore R converts character class to integer class. This may cause the misleading understanding. But, now we have ds_a$y double class and ds_b$y integer class -> but this is no problem for R and bind_rows here double class overrides integer.

> # showing what type.convert does to ds_b
> ds_b$x <- as.integer(ds_b$x)
> ds_b$y <- as.integer(ds_b$y)
> ds_b %>% 
+   as_tibble()
# A tibble: 6 x 3
      x     y     p
  <int> <int> <dbl>
1     1     5     2
2     2     5     2
3     3     5     2
4     4     5     2
5     5     5     2
6     6     5     2
> ds_b %>% 
+   as_tibble()
# A tibble: 6 x 3
      x     y     p
  <int> <int> <dbl>
1     1     5     2
2     2     5     2
3     3     5     2
4     4     5     2
5     5     5     2
6     6     5     2
> bind_rows(ds_a, ds_b) %>% 
+   as_tibble()
# A tibble: 12 x 5
       x     y z         l     p
   <int> <dbl> <chr> <dbl> <dbl>
 1     1     5 4         2    NA
 2     2     5 4         2    NA
 3     3     5 4         2    NA
 4     4     5 4         2    NA
 5     5     5 4         2    NA
 6     6     5 4         2    NA
 7     1     5 NA       NA     2
 8     2     5 NA       NA     2
 9     3     5 NA       NA     2
10     4     5 NA       NA     2
11     5     5 NA       NA     2
12     6     5 NA       NA     2

3. converts ds_b$p which is class double to class integer because the data are integer in nature.

Solution:

library(dplyr)
bind_rows(ds_a, ds_b %>% type.convert(as.is=TRUE))

output:

   x y    z  l  p
1  1 5    4  2 NA
2  2 5    4  2 NA
3  3 5    4  2 NA
4  4 5    4  2 NA
5  5 5    4  2 NA
6  6 5    4  2 NA
7  1 5 <NA> NA  2
8  2 5 <NA> NA  2
9  3 5 <NA> NA  2
10 4 5 <NA> NA  2
11 5 5 <NA> NA  2
12 6 5 <NA> NA  2

Answer 2

You can change the class of one dataframe according to another one and row bind the datasets.

library(dplyr)
library(purrr)

bind_rows(ds_a, map2_df(ds_b, map(ds_a, class), ~{class(.x) <- .y;.x}))

#   x y
#1  1 5
#2  2 5
#3  3 5
#4  4 5
#5  5 5
#6  6 5
#7  1 5
#8  2 5
#9  3 5
#10 4 5
#11 5 5
#12 6 5

map2_df is used to changes the class of ds_b data where

.x - passes the column value of ds_b.

.y - map(ds_a, class) gets the class of each column in ds_a

In the function it changes class of .x with .y value and bind them. We then use bind_rows with ds_a dataframe.

---

If there are unequal number of columns you can change the classes of only common ones and bind the rows.

new_bind <- function(a, b) {
  common_cols <- intersect(names(a), names(b))
  b[common_cols] <- map2_df(b[common_cols], 
               map(a[common_cols], class), ~{class(.x) <- .y;.x})
  bind_rows(a, b)  
}
new_bind(ds_a, ds_b) 

#   x y    z  l  p
#1  1 5    4  2 NA
#2  2 5    4  2 NA
#3  3 5    4  2 NA
#4  4 5    4  2 NA
#5  5 5    4  2 NA
#6  6 5    4  2 NA
#7  1 5 <NA> NA  2
#8  2 5 <NA> NA  2
#9  3 5 <NA> NA  2
#10 4 5 <NA> NA  2
#11 5 5 <NA> NA  2
#12 6 5 <NA> NA  2