do {
    // initialization failed, looks like I can not use "\\" here
    let regex = try NSRegularExpression.init(pattern: "(?<!\\)\n")

    let string = """
    aaabbb
    zzz
    """
    
    // expect "aaabbb\nzzz"
    print(regex.stringByReplacingMatches(in: string, options: [], range: NSMakeRange(0, string.count), withTemplate: "\\n"))
} catch let error {
    print(error)
}

Here I want to replace "\n" in my string with "\\n", but failed at the very beginning, the error message is

// NSRegularExpression did not recognize the pattern correctly.
Error Domain=NSCocoaErrorDomain Code=2048 "The value “(?<!\)
” is invalid." UserInfo={NSInvalidValue=(?<!\)
}

The regex has been tested in regular expression 101, so it is right, just doesn't work in Swift for some reason.

How can I do this?

CodePudding user response：

Base on Larme's comment:

in Swift, \ (double back slash) in a String is for "having a ``, as you see in the error, you have (?<!\), but it means then that you are escaping the closing ), so you have a missing closing ). I'd say that you should write then "(?<!\\\\)\n"?

I finally figured out what's going on and how to fix it.

The problem is backslash.

In Swift, a backslash inside double quotation mark would be treated as escape sequence, like this

// won't compile
// error: Invalid escape sequence in literal
let regex = try NSRegularExpression.init(pattern: "(?<!\)\n")

If we add another backslash, is it work?

No, cause these 2 backslashes would be treated as a single escape character for the upcoming closing ).

// compile but get a runtime error
let regex = try NSRegularExpression.init(pattern: "(?<!\\)\n")

Hence the runtime error

NSRegularExpression did not recognize the pattern correctly.
Error Domain=NSCocoaErrorDomain Code=2048 "The value “(?<!\)
” is invalid." UserInfo={NSInvalidValue=(?<!\)

To show that what we need is a literal backslash, we actually need 4 backslashes

let regex = try NSRegularExpression.init(pattern: "(?<!\\\\)\n")

The first two backslashes represent an escape character and the last two represent one literal backslash.

These seem very troublesome.

Better Approach

Fortunately, starting with Swift 5, we can use a pair of # to do this

// works like in online tool
let regex = try NSRegularExpression.init(pattern: "(?<!\\)\n")

Another thing

It’s worth noticing that the initialization of regular expression isn’t the only thing that requires special handling

// withTemplate
print(regex.stringByReplacingMatches(in: string, options: [], range: NSMakeRange(0, string.count), withTemplate: #"\\n"#))

// As a comparison, this is OK
print(string.replacingOccurrences(of: "\n", with: "\\N"))