package MultibyteDiagnostics
import ChunkedString
import ClosureTimers

/*	In-game diagnostics for multibyte (non-latin) string support.

	The behavior of SubString/StringHash for slices of multibyte characters is
	undocumented and has changed between game versions, and the wurst interpreter
	(unit tests) does not emulate it. The only way to verify that multibyte
	support works on the current patch is to check in game.

	Configure the stdlib with `ENABLE_MULTIBYTE_SUPPORT = true`, then call
	`runMultibyteDiagnostics()` from your map (e.g. via a chat command) and
	inspect the output. Lines prefixed with [FAIL] indicate that multibyte
	support is broken on the current game version, in which case non-latin
	characters (player names, chat input) are not safe to store in save data.

	Observed semantics on 2.0 era patches:
	- SubString is byte-indexed, raw and lossless - slices keep their bytes
	  even when they cut a character in half.
	- StringHash collapses slices starting with the lead byte of a 2-3 byte
	  character to a constant marker hash (PARTIAL_CHAR_HASH).
	- Slices of a single continuation byte and slices starting with a 4-byte
	  lead byte (emoji) are NOT collapsed and keep per-content hashes.
*/

function check(string name, boolean condition)
	print((condition ? "[ok] " : "|cffff0000[FAIL]|r ") + name)

public function runMultibyteDiagnostics()
	print("ENABLE_MULTIBYTE_SUPPORT = " + ENABLE_MULTIBYTE_SUPPORT.toString())
	print("PARTIAL_CHAR_HASH = " + PARTIAL_CHAR_HASH.toString() + " (legacy = 1843378377)")
	print("LEAD4_HASH = " + LEAD4_HASH.toString())

	// Raw engine semantics: 2-byte (cyrillic/umlauts), 3-byte (CJK), 4-byte (emoji)
	let two = "дä"
	let three = "中文"
	let four = "😀"
	check("2-byte char has length 2", "д".length() == 2)
	check("3-byte char has length 3", "中".length() == 3)
	check("4-byte char has length 4", four.length() == 4)
	check("partial 2-byte slice is detected", "д".substring(0, 1).isMultibytePartial())
	check("partial 3-byte slices are detected", "中".substring(0, 1).isMultibytePartial() and "中".substring(0, 2).isMultibytePartial())
	check("full char slice is valid", two.substring(0, 2) == "д")
	check("slices are lossless", two.substring(0, 1) + two.substring(1, 4) == two)
	// 4-byte lead slices are expected to keep their own hash instead of the marker
	print("4-byte slice hashes: " + four.substring(0, 1).getHash().toString() + ", "
		+ four.substring(0, 2).getHash().toString() + ", " + four.substring(0, 3).getHash().toString())
	check("4-byte lead slice matches LEAD4_HASH", four.substring(0, 1).getHash() == LEAD4_HASH)

	// Boundary classification
	check("boundary detected at start of char", two.isCharBoundary(2))
	check("boundary rejected inside 2-byte char", not two.isCharBoundary(1) and not two.isCharBoundary(3))
	check("boundary rejected inside 3-byte char", not three.isCharBoundary(1) and not three.isCharBoundary(2))
	check("boundary detected between 3-byte chars", three.isCharBoundary(3))
	check("boundary rejected inside 4-byte char", not four.isCharBoundary(1) and not four.isCharBoundary(2) and not four.isCharBoundary(3))
	check("boundary detected before 4-byte char", ("a" + four).isCharBoundary(1))

	// Character-wise iteration
	var iterated = ""
	var count = 0
	for c in "aд中😀b"
		iterated += c
		count++
	check("iteration reassembles string", iterated == "aд中😀b")
	check("iteration counts 5 chars, got " + count.toString(), count == 5)

	// ChunkedString round trips with cuts landing inside multibyte chars
	let input = "абвгдеёжз中文漢字😀🙂ab"
	for chunkSize = 3 to 8
		let cstring = new ChunkedString(chunkSize)
		cstring.append(input)
		check("ChunkedString round trip, chunkSize " + chunkSize.toString(), cstring.getUnsafeString() == input)
		destroy cstring