tidyr separate column values into character and numeric using regex

Question

I'd like to separate column values using tidyr::separate and a regex expression but am new to regex expressions

df <- data.frame(A=c("enc0","enc10","enc25","enc100","harab0","harab25","harab100","requi0","requi25","requi100"), stringsAsFactors=F) 

This is what I've tried

library(tidyr)
df %>%
   separate(A, c("name","value"), sep="[a-z]+")

Bad Output

   name value
1           0
2          10
3          25
4         100
5           0
# etc

How do I save the name column as well?

Answer 1

You may use a (?<=[a-z])(?=[0-9]) lookaround based regex with tidyr::separate:

> tidyr::separate(df, A, into = c("name", "value"), "(?<=[a-z])(?=[0-9])")
    name value
1    enc     0
2    enc    10
3    enc    25
4    enc   100
5  harab     0
6  harab    25
7  harab   100
8  requi     0
9  requi    25
10 requi   100

The (?<=[a-z])(?=[0-9]) pattern matches a location in the string right in between a lowercase ASCII letter ((?<=[a-z])) and a digit ((?=[0-9])). The (?<=...) is a positive lookahead that requires the presence of some pattern immediately to the left of the current location, and (?=...) is a positive lookahead that requires the presence of its pattern immediately to the right of the current location. Thus, the letters and digits are kept intact when splitting.

Alternatively, you may use extract:

extract(df, A, into = c("name", "value"), "^([a-z]+)(\\d+)$")

Output:

    name value
1    enc     0
2    enc    10
3    enc    25
4    enc   100
5  harab     0
6  harab    25
7  harab   100
8  requi     0
9  requi    25
10 requi   100

The ^([a-z]+)(\\d+)$ pattern matches:

• ^ - start of input
• ([a-z]+) - Capturing group 1 (column name): one or more lowercase ASCII letters
• (\\d+) - Capturing group 2 (column value): one or more digits
• $ - end of string.

Answer 2

For a bare R version without a lookaround-based regex, define the regular expression first:

> re <- "[a-zA-Z][0-9]"

Then use two substr() commands to separate and return the desired two components, before and after the matched pattern.

> with(df,
      data.frame(name=substr(A, 1L, regexpr(re, A)), 
                 value=substr(A, regexpr(re, A) + 1L, 1000L))
      )
    name value
1    enc     0
2    enc    10
3    enc    25
4    enc   100
5  harab     0
6  harab    25
7  harab   100
8  requi     0
9  requi    25
10 requi   100

The regex here looks for the pattern "any alpha" [a-zA-Z] followed by "any numeric" [0-9]. I believe this is what the reshape command does if the sep argument is specified as "".

Answer 3

You can add one more step If you really want to get it with separate, in which I don't see the point, i.e. (Using the same regex as @ WiktorStribiżew),

df %>% 
  mutate(A = gsub('^([a-z]+)(\\d+)$', '\\1_\\2', A)) %>% 
  separate(A, into = c('name', 'value'), sep = '_')