Case Folding and Normalization

utf8UncasedFind(substring:) performs a case-insensitive search using full Unicode folding and returns a byte-accurate Range<Index>?. The matched length can differ from the needle length, since folding can change byte counts.

if let range = "Grüße".utf8UncasedFind(substring: "GRÜSSE") {
    print("matched", "Grüße"[range])
}

For repeated case-insensitive searches with the same needle, build a Utf8UncasedNeedle once and reuse it. The needle is prepared once and never changes, so one needle can search from many threads at once. An empty needle matches at the start of any haystack with an empty range.

let needle = Utf8UncasedNeedle("hello")
let r = needle.findFirst(in: "Say HELLO to the world")
assert(r != nil)

utf8UncasedFoldedBytes() returns the fully case-folded UTF-8 bytes, which may be longer than the input since "ß" folds to "ss".

let folded = "Straße".utf8UncasedFoldedBytes()  // [UInt8] of "strasse"
assert(folded == Array("strasse".utf8))

Normalization is driven by StringZillaNormalizationForm, one of .nfc, .nfd, .nfkc, or .nfkd. utf8Normalized(_:) returns the normalized UTF-8 bytes and defaults to .nfc, utf8NormalizationViolation(_:) returns the Index? of the first non-conforming byte, and isUtf8Normalized(_:) is a convenience Bool.

let nfc = "e\u{0301}".utf8Normalized(.nfc)  // composed "é" bytes
assert("café".isUtf8Normalized(.nfc))       // already composed
let fi = "\u{FB01}".utf8Normalized(.nfkc)   // ligature "fi" → "fi" bytes
assert(fi == Array("fi".utf8))
let bad = "e\u{0301}".utf8NormalizationViolation(.nfc) // Index of the violation
assert(bad != nil)