Order() function in r is not ordering correctly

Question

I have a df of numbers and am doing some ordering. The output is placing 7 next to 70 as if 7 is 70. Why is this happening. The pasted stuff below is the actual output. Notice how 263 is treated smaller than 27 as if there is a 0 behind the 7 in 27. 4 is after 38 as if 4 means 40. I'm using the order().

 feat_1  25
 feat_10  26
 feat_24 263
 feat_48  27
 feat_55  27
 feat_75  36
 feat_16  37
 feat_53  38
 feat_89  38
 feat_28   4

Answer 1

This is happening because you are sorting characters instead of number. It is a common problem, although not a visible one. For starters, it's easy to use order to sort a data.frame, that's what I'll be using to demonstrate a solution in my test case.

You should try this:

col1 <- c('a', 'b', 'c')
col2 <- c("25", "42" ,"4")
df <- data.frame(col1, col2)

## This is the wrong approach:
df[order(df$col2),]
col1 col2
1   a   25
3   c    4
2   b   42

## This is the right approach, conver the second vector to numeric vector:
df$col2 <- as.numeric(as.character(df$col2))
df[order(df$col2),]
  col1 col2
3   c    4
1   a   25
2   b   42

Answer 2

You could also use mixedsort or mixedorder from the gtools package (for a fast alternative) and there is no need to convert the column to numeric because it deals with either character numbers or alphanumeric strings:

Data

df <- read.table(text='feat_1  25
 feat_10  "26"
 feat_24  "263"
 feat_48  "27"
 feat_55  "27"
 feat_75  "36"
 feat_16  "37"
 feat_53  "38"
 feat_89  "38"
 feat_28   "4"')

Solution

library(gtools)
#you use mixedorder in exactly the same way as base order
> df[mixedorder(df$V2),]
        V1  V2
10 feat_28   4
1   feat_1  25
2  feat_10  26
4  feat_48  27
5  feat_55  27
6  feat_75  36
7  feat_16  37
8  feat_53  38
9  feat_89  38
3  feat_24 263