Swift Regex matching fails when source contains unicode characters

Question

I'm trying to do a simple regex match using NSRegularExpression, but I'm having some problems matching the string when the source contains multibyte characters:

let string = "D 9"

// The following matches (any characters)(SPACE)(numbers)(any characters)
let pattern = "([\\s\\S]*) ([0-9]*)(.*)"

let slen : Int = string.lengthOfBytesUsingEncoding(NSUTF8StringEncoding)

var error: NSError? = nil

var regex = NSRegularExpression(pattern: pattern, options: NSRegularExpressionOptions.DotMatchesLineSeparators, error: &error)

var result = regex?.stringByReplacingMatchesInString(string, options: nil, range: NSRange(location:0,
length:slen), withTemplate: "First \"$1\" Second: \"$2\"")

The code above returns "D" and "9" as expected

If I now change the first line to include a UK 'Pound' currency symbol as follows:

let string = "£ 9"

Then the match doesn't work, even though the ([\\s\\S]*) part of the expression should still match any leading characters.

I understand that the £ symbol will take two bytes but the wildcard leading match should ignore those shouldn't it?

Can anyone explain what is going on here please?

I'm not familiar with Swift and its regex engine, but in general I would be terribly surprised to find that `\s\S` isn't equivalent to `.` when Unicode is involved. Why aren't you using `.*` in the first grouping? That said, I'm not entirely convinced that tha's where the problem is, either; I think it's more likely that `[0-9]` fails to match unicode digits than that `\S` fails to match arbitrary non-space unicode characters. — Kyle Strand, Apr 20 '15 at 19:23
Swift *does* support the `\d` character class, so why are you using `[0-9]`? If you try matching with `(.*) (\d*)(.*)`, do you get a match? — Kyle Strand, Apr 20 '15 at 19:27
Thanks Kyle. I was using \s\S because of a mis-reading of an article about misuse of the '.' character. I've changed it to "(.*) (\d*)(.*)" but it still fails to match. I'm beginning to suspect it is a bug in the Swift implementation - any other character matches OK - E.g. "D$+@ 9" but when I put a '£' symbol anywhere in the string to be matched, it fails! — NEIL STRONG, Apr 22 '15 at 05:54

Martin R · Accepted Answer · 2015-04-22T05:58:03.007

It can be confusing. The first parameter of stringByReplacingMatchesInString() is mapped from NSString in Objective-C to String in Swift, but the range: parameter is still an NSRange. Therefore you have to specify the range in the units used by NSString (which is the number of UTF-16 code points):

var result = regex?.stringByReplacingMatchesInString(string,
        options: nil,
        range: NSRange(location:0, length:(string as NSString).length),
        withTemplate: "First \"$1\" Second: \"$2\"")

Alternatively you can use count(string.utf16) instead of (string as NSString).length .

Full example:

let string = "£ 9"

let pattern = "([\\s\\S]*) ([0-9]*)(.*)"
var error: NSError? = nil
let regex = NSRegularExpression(pattern: pattern,
        options: NSRegularExpressionOptions.DotMatchesLineSeparators,
        error: &error)!

let result = regex.stringByReplacingMatchesInString(string,
    options: nil,
    range: NSRange(location:0, length:(string as NSString).length),
    withTemplate: "First \"$1\" Second: \"$2\"")
println(result)
// First "£" Second: "9"

Thank you Martin - that does explain why the length of the string containing the currency symbol was being reported as 4 rather than 3. It reports the length correctly now, but the expression still doesn't match I'm afraid. — NEIL STRONG, Apr 22 '15 at 05:47
My apologies Martin - must be too early in the morning - your solution DID work!! Many thanks!! :-) — NEIL STRONG, Apr 22 '15 at 06:08
Thank you soo much! I spent so much time trying to resolve this. — 3li, Apr 06 '20 at 15:12

score 0 · Answer 2 · answered Jun 01 '22 at 18:33

I've run into this a couple times and Martin's answer helped me understand the problem. Here's a quick version of the solution that worked for me.

If your regular expression function includes a range parameter built like this:

NSRange(location: 0, length: yourString.count)

You can change it to this:

NSRange(location: 0, length: yourString.utf16.count)

Swift Regex matching fails when source contains unicode characters

2 Answers2

Linked