（提取/分离/匹配）组中的任何顺序

Question

# Sample Data Frame
df  <- data.frame(Column_A 
                  =c("1011 Red Cat", 
                     "Mouse 2011 is in the House 3001", "Yellow on Blue Dog walked around Park"))

I've a column of manually inputted data which I'm trying to clean.

  Column_A 
1|1011 Red Cat                         |
2|Mouse 2011 is in the House 3001      |
2|Yellow on Blue Dog walked around Park|  

I want to separate each characteristic into it's own column, but still maintain Column A to pull out other characteristics later.

  Colour               Code           Column_A
1|Red                 |1001          |Cat
2|NA                  |2001 3001     |Mouse is in the House
3|Yellow on Blue      |NA            |Dog walked around Park

To date, I've been re-ordering them with gsub and capturing groups, then using Tidyr::extract to separate them.

library(dplyr)
library(tidyr)
library(stringr)

df1 <- df %>% 

  # Reorders the Colours
  mutate(Column_A = gsub("(.*?)?(Yellow|Blue|Red)(.*)?", "\\2 \\1\\3", 
                         Column_A, perl = TRUE)) %>%
  # Removes Whitespaces 
  mutate(Column_A =str_squish(Column_A)) %>%
  # Extracts the Colours 
  extract(Column_A, c("Colour", "Column_A"), "(Red|Yellow|Blue)?(.*)") %>%

  # Repeats the Prececding Steps for Codes
  mutate(Column_A = gsub("(.*?)?(\\b\\d{1,}\\b)(.*)?", "\\2 \\1\\3", 
                         Column_A, perl = TRUE)) %>%
  mutate(Column_A =str_squish(Column_A)) %>%
  extract(Column_A, c("Code", "Column_A"), "(\\b\\d{1,}\\b)?(.*)") %>%
  mutate(Column_A = str_squish(Column_A))

Which Results in this:

Colour      Code    Column_A
|Red        |1011   |Cat
|Yellow     |NA     |on Blue Dog walked around Park
|NA         |1011   |Mouse is in the House 1001

This works fine for the first row, but not the proceeding space and word separated ones, which I've subsequently been extracting and uniting. What's a more elegant way of doing this?

Answer 1

Here's a solution with a combination of stringr and gsub , using a list of colours supplied in R:

library(dplyr)
library(stringr)

# list of colours from R colors()
cols <- as.character(colors())

apply(df,
      1,
      function(x)

        tibble(
          # Exctract CSV of colours
          Color = cols[cols %in% str_split(tolower(x), " ", simplify = T)] %>%
            paste0(collapse = ","),

          # Extract CSV of sequential lists of digits
          Code = str_extract_all(x, regex("\\d+"), simplify = T) %>%
            paste0(collapse = ","),

          # Remove colours and digits from Column_A
          Column_A = gsub(paste0("(\\d+|",
                                 paste0(cols, collapse = "|"),
                                 ")"), "", x, ignore.case = T) %>% trimws())) %>%
  bind_rows()

# A tibble: 3 x 3
  Color       Code      Column_A                  
  <chr>       <chr>     <chr>                     
1 red         1011      Cat                       
2 ""          2011,3001 Mouse  is in the House    
3 blue,yellow ""        on  Dog walked around Park

Answer 2

Using tidyverse we can do

library(tidyverse)

colors <- paste0(c("Red", "Yellow", "Blue"), collapse = "|")

df %>%
   mutate(Color = str_extract(Column_A,
                   paste0("(", colors, ").*(", colors, ")|(", colors, ")")),
           Code = str_extract_all(Column_A, "\\d+", ), 
           Column_A = pmap_chr(list(Color, Code, Column_A), function(x, y, z) 
              trimws(gsub(paste0("\\b", c(x,  y), "\\b", collapse = "|"), "", z))), 
           Code = map_chr(Code, paste, collapse = " "))


#                 Column_A         Color      Code
#1                    Cat            Red      1011
#2 Mouse  is in the House           <NA> 2011 3001
#3 Dog walked around Park Yellow on Blue      

We first extract text between two colors using str_extract . You can include all the possible colors which can occur in the data in colors . We use paste0 to construct the regex. For this example it would be

paste0("(", colors, ").*(", colors, ")|(", colors, ")")
#[1] "(Red|Yellow|Blue).*(Red|Yellow|Blue)|(Red|Yellow|Blue)"

meaning extract text between and including colors or extract only colors .

For Code part as we can have multiple Code values, we use str_extract_all and get all the numbers from the column. This part is initially stored in a list.

For Column_A values we remove everything which was selected in Code and Color adding word boundaries using gsub and the remaining part is saved.

As we had stored Code in list previously, we convert them to one string by collapsing them. This returns empty strings for values that do not match. You can convert them back to NA by adding Code = replace(Code, Code == "", NA)) in the chain if needed.