Mask phone number in R-CodePudding

My raw data has a lot of personal information, so I am masking them in R. The sample data and my original code are below:

install.packages("stringr")
library(string)

x = c("010-1234-5678",
      "John 010-8888-8888",
      "Phone: 010-1111-2222",
      "Peter 018.1111.3333",
      "Year(2007,2019,2020)",
      "Alice 01077776666")

df = data.frame(
  phoneNumber = x
)

pattern1 = "\\d{3}-\\d{4}-\\d{4}"
pattern2 = "\\d{3}.\\d{4}.\\d{4}"
pattern3 = "\\d{11}"

delPhoneList1 <- str_match_all(df, pattern1) %>% unlist
delPhoneList2 <- str_match_all(df, pattern2) %>% unlist
delPhoneList3 <- str_match_all(df, pattern3) %>% unlist

I found three types of patterns from the dataset and each result is below:

> delPhoneList1
[1] "010-1234-5678" "010-8888-8888" "010-1111-2222"
> delPhoneList2
[1] "010-1234-5678" "010-8888-8888" "010-1111-2222" "018.1111.3333" "007,2019,2020"
> delPhoneList3
[1] "01077776666"

Pattern1 is the typical type of phone number in my country using a dash, but someone types in the number like pattern2 using a comma. However, pattern2 also includes pattern1, so it detects the other pattern like a series of the year. It is an unexpected result.

My question is how to match the exact pattern that I define. The pattern2 includes excessive patterns such as "007,2019,2020" from "Year(2007,2019,2020)".

Additionally, the next step is masking the number using the below code:

for (phone in delPhoneList1) {
  df$phoneNumber <- gsub(phone, "010-9999-9999", df$phoneNumber)
}

I think the code is perfect for me, but if you had a more efficient way, please let me know.

Thanks.

CodePudding user response：

One pattern to rule them all ;-)

ptn <- "\\b\\d{3}([-.]?)\\d{4}\\1\\d{4}\\b"
grepl(ptn, x)
# [1]  TRUE  TRUE  TRUE  TRUE FALSE  TRUE

The reason your pattern2 failed is because it used . as a separator, but in regex that means "any character". You could have use \\. instead of . and it would have behaved better.
I'm using place holders here: if the first separator is a -, then \\1 ensures that the other separator is the same. If it's empty, then the second is empty as well. This also allows the 11 uninterrupted numbers of pattern3.
The \\b are word-boundaries, assuring us that 12-digits would not match:
```
grepl(ptn, c("12345678901", "123456789012"))
# [1]  TRUE FALSE
```

Since this has a placeholder, it tends to mess a little with stringr:: functions, but we can work around that, depending on what you need.

For instance, if you replace the placeholder with a second instance of the same pattern, it might allow 123-4444.5555 (mixed separators), if that's not a problem.

ptn2 <- "\\b\\d{3}[-.]?\\d{4}[-.]?\\d{4}\\b"
unlist(str_match_all(x, ptn2))
# [1] "010-1234-5678" "010-8888-8888" "010-1111-2222" "018.1111.3333" "01077776666"

or we can exploit the number of patterns matched (original ptn):

unlist(str_match(x, ptn)[,1])
# [1] "010-1234-5678" "010-8888-8888" "010-1111-2222" "018.1111.3333" NA              "01077776666"