R: Passing Column Names to Function w/ dplyr

Question

I am trying to set up a function in R which prepares data in a specific format to be fed into a correlogram. When manipulating datasets I tend to use dplyr due to its clarity and ease of use, but I am running into problems trying to pass a dataset and specified column names into this function while using dplyr.

This is the set up, included here (in slightly abbreviated form) for clarity. I have not encountered any errors with this and before posting this I confirmed corrData is set up properly:

library(corrplot)
library(tidyverse)
library(stringr)

table2a <- table2 %>%
    mutate(example_index = str_c(country,year, sep="."))

Here is the actual function:

prepCorr <- function(dtable, x2, index2) { 

  practice <- dtable %>%
    select(index2, x2) %>%
    mutate(count=1) %>%
    complete(index2, x2)

  practice$count[is.na(practice$count)] <- 0

  practice <- spread(practice, key = x2, value = count)

  M <- cor(practice)
  return(M)
}

prepCorr(table2a, type, example_index)

Whenever I run this function I get:

Error in overscope_eval_next(overscope, expr) : object 'example_index' not found

I have also tried to take advantage of quosures to fix this, but recieve a different error when I do so. When I run the following modified code:

prepCorr <- function(dtable, x2, index2) { 
  x2 <- enquo(x2)
  index2 <- enquo(index2)

  practice <- dtable %>%
    select(!!index2, !!x2) %>%
    mutate(count=1) %>%
    complete(!!index2, !!x2)

  practice$count[is.na(practice$count)] <- 0

  practice <- spread(practice, key = !!x2, value = count)

  return(cor(practice))
}

prepCorr(table2a, type, example_index)

I get:

Error in !index2 : invalid argument type

What am I doing wrong here, and how can I fix this? I believe I am using dplyr 0.7 for clarification.

UPDATE: replaced old example with reproducible example.

Can you post a [reproducible example](https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example)? — csgroen, Aug 22 '17 at 20:05
Would it be alright if I edited the question to use table2 from the tidyverse package rather than the tables shown here? It should work for this purpose nicely. — Josh Kraushaar, Aug 22 '17 at 20:32
Sure, as long as it gives the same error and we can use it to "debug" — csgroen, Aug 22 '17 at 20:34
Ok, it should be reproducible now, thank you for the pointer. — Josh Kraushaar, Aug 22 '17 at 20:48
Save the RStudio for RStudio-speific problems. (The RStudio interface, or code that works in RGui or on the R command line, but not in RStudio.) — Gregor Thomas, Aug 22 '17 at 21:24
Your example with quosures works for me (with dplyr 0.7.2 and tidyr 0.7.0). — aosmith, Aug 23 '17 at 15:39
Any suggestions as to why this might be encountering issues? It seems like it is processing !!index2 as !(!index2). Can I avoid that? — Josh Kraushaar, Aug 25 '17 at 20:22

score 2 · Answer 1 · answered Aug 22 '17 at 20:31

2

Look at this example

library(dplyr)
myfun <- function(df, col1, col2) {
              col1 <- enquo(col1)     # need to quote
              col2 <- enquo(col2)
              df1 <- df %>%
                       select(!!col1, !!col2)   #!! unquotes
              return(df1)
         }

myfun(mtcars, cyl, gear)

You can learn more here link about NSE vs SE

answered Aug 22 '17 at 20:31

CPak

13,260
3
30
48

When I run that, using say index2 <- enquo(index2) I keep getting: ** "Error in !index2 : invalid argument type'" ** – Josh Kraushaar Aug 22 '17 at 20:39
I am using double exclamation marks, but weirdly the error only returns !index2 rather than !!index2 – Josh Kraushaar Aug 22 '17 at 20:45
Looks like it's failing on `tidyr::complete`. You should ask a separate question to see if NSE evaluations are possible with `tidyr::complete` and `tidyr::spread` – CPak Aug 22 '17 at 21:04
1

It fails if I just use tidyr::select, and it fails before tidyr::spread is even processed (or if I comment tidyr::spread out and only test the first block of tidyr code) – Josh Kraushaar Aug 22 '17 at 21:10
Use `dplyr::select` instead – CPak Aug 22 '17 at 21:13
1

It seems like it is processing !!index2 as !(!index2). Can I avoid that? – Josh Kraushaar Aug 25 '17 at 20:23

R: Passing Column Names to Function w/ dplyr

1 Answers1