Filter data frame inside a function with a column argument

Question

I'm trying to update a data frame column inside a function based on a filtered column.

#example dataframe
my.df = data.frame(A=1:10)

#define function to classify column passed as argument 2 based on argument 3
classify = function(df, col, threshold){
  df[df$col<threshold, 2] <- "low"
  df[df$col>=threshold, 2] <- "high"

  return(df)
}

#assign output to new.df
new.df = classify(my.df, A, 5)

I'd expect the new column to contain character values of 'low' or 'high', but instead they're all <NA>.

mutate(my.df, B = ifelse(A < 2, 'low', 'high')) ? – Sean Lin May 12 '17 at 19:34 — Sean Lin, May 12 '17 at 19:34

score 1 · Accepted Answer · answered May 12 '17 at 19:56

Simply pass the string literal of the column name, "A", and then inside the function receive the parameter with single or double bracket [[...]] indexing and not with $:

# example dataframe
my.df = data.frame(A=1:10)

# define function to classify column passed as argument 2 based on argument 3
classify = function(df, col, threshold){
  df[df[[col]] < threshold, 2] <- "low"
  df[df[[col]] >= threshold, 2] <- "high"

  return(df)
}

# assign output to new.df
new.df = classify(my.df, "A", 5)

new.df    
#     A   V2
# 1   1  low
# 2   2  low
# 3   3  low
# 4   4  low
# 5   5 high
# 6   6 high
# 7   7 high
# 8   8 high
# 9   9 high
# 10 10 high

score 0 · Answer 2 · answered May 12 '17 at 19:34

We can use the devel version of dplyr (soon to be released 0.6.0) to do this. The enquo takes the input argument and converts it to quosure which will be evaluated within the mutate/group_by/filter etc by unquoting (UQ)

library(dplyr)
classify <- function(df, col, threshold){
   col <- enquo(col)

   df %>%
       mutate(categ = ifelse(UQ(col) < threshold, "low", "high"))

}

classify(my.df, A, 5)
#    A categ
#1   1   low
#2   2   low
#3   3   low
#4   4   low
#5   5  high
#6   6  high
#7   7  high
#8   8  high
#9   9  high
#10 10  high

Filter data frame inside a function with a column argument

2 Answers2