Wsrtjtyk

I think the title is rather straight forward.
But just to provide some data and example:

test <- tibble(

ID1 = letters,

ID2 = LETTERS,

A1 = runif(26),

B1 = runif(26),

A2 = runif(26),

B2 = runif(26)

)

Is there a way to gather on only e.g., numerical columns with a simple command such that:

test %>% gather_if(is.numeric, 'key', 'value')

? Which would give the same output as the following:

> test %>% gather('key', 'value', -ID1, -ID2)

# A tibble: 104 x 4

   ID1   ID2   key    value

   <chr> <chr> <chr>  <dbl>

 1 a     A     A1    0.558 

 2 b     B     A1    0.0614

 3 c     C     A1    0.999 

 4 d     D     A1    0.854 

 5 e     E     A1    0.463 

 6 f     F     A1    0.875 

 7 g     G     A1    0.796 

 8 h     H     A1    0.484 

 9 i     I     A1    0.336 

10 j     J     A1    0.191 

# ... with 94 more rows

Looking at the gather function:

> gather

function (data, key = "key", value = "value", ..., na.rm = FALSE, 

    convert = FALSE, factor_key = FALSE) 

{

    UseMethod("gather")

}

<bytecode: 0x000000001b71ff18>

<environment: namespace:tidyr>

It does not seem that straight forward to modify it (at least not for me who is a semi-novel R user).

Edit:

My vocabulary choice in dplyr might not be exactly accurate. But I think the MWE explains fairly well what type of function I'm going for.

Edit2:

Using the answer by bschneidr, an ad-hoc version of this could be done in the following way.

gather_if <- function(data, fun, key, value, ..., na.rm = FALSE, convert = 

FALSE, factor_key = FALSE){

    data %>%

        gather(!!key, !!value, select_if(., fun) %>% colnames(), ...,

               na.rm = FALSE, convert = FALSE, factor_key = FALSE)

}

Which gives:

> test %>% gather_if(is.numeric, 'key', 'value')

# A tibble: 104 x 4

   ID1   ID2   key    value

   <chr> <chr> <chr>  <dbl>

 1 a     A     A1    0.558 

 2 b     B     A1    0.0614

 3 c     C     A1    0.999 

 4 d     D     A1    0.854 

 5 e     E     A1    0.463 

 6 f     F     A1    0.875 

 7 g     G     A1    0.796 

 8 h     H     A1    0.484 

 9 i     I     A1    0.336 

10 j     J     A1    0.191 

# ... with 94 more rows

I think a gather_if function is in the works for tidyr (see this pull request on tidyr's Github repo).

For the moment, I think the easiest method is to use dplyr's select_if function inside the call to gather.

test %>% 

    gather('key', 'value',

           colnames(select_if(., is.numeric)))

One way to do it is to negate() the numeric condition and extract the names. It does look a bit cumbersome but here it is,

library(tidyverse)



gather(test, key, value, -c(test %>% select_if(negate(is.numeric)) %>% names()))

which gives,

# A tibble: 104 x 4

   ID1   ID2   key    value

   <chr> <chr> <chr>  <dbl>

 1 a     A     A1    0.624 

 2 b     B     A1    0.0740

 3 c     C     A1    0.790 

 4 d     D     A1    0.312 

 5 e     E     A1    0.323 

 6 f     F     A1    0.826 

 7 g     G     A1    0.0533

 8 h     H     A1    0.0828

 9 i     I     A1    0.979 

10 j     J     A1    0.453 

# ... with 94 more rows

If think you want this:

gather_if <- function(data, FUN, key = "key", value = "value", na.rm = FALSE, convert = FALSE, factor_key = FALSE) {

    data %>% {gather(., key = key, value = value , names(.)[sapply(., FUN = FUN)], na.rm = na.rm, convert = convert, factor_key = factor_key )}

} 

call your new cool function:

test %>% gather_if(is.numeric, 'key', 'value')

result:

# A tibble: 104 x 4

#   ID1   ID2   key   value

#   <chr> <chr> <chr> <dbl>

# 1 a     A     A1    0.693

# 2 b     B     A1    0.356

# 3 c     C     A1    0.650

# 4 d     D     A1    0.358

# 5 e     E     A1    0.650

# 6 f     F     A1    0.461

# 7 g     G     A1    0.222

# 8 h     H     A1    0.993

# 9 i     I     A1    0.679

#10 j     J     A1    0.331

# ... with 94 more rows