Alba Mendez mildsunrise

[lz-string][] is a very popular library that compresses UTF-16 strings using a variation of the [LZ78][] algorithm where literals are only encoded once (and referred as dictionary indexes afterwards).

Despite its impressive 10 million downloads per week at the time of this writing, there is no official documentation on the wire format implemented by this library, and the horrible code quality makes it hard to understand from it. Furthermore, the code contains many bugs / gotchas, some of which are enumerated below. This is probably why many of the [numerous ports][ports] of the library blindly copy its code, doing little more

JNIEnv

(keep in mind that JNI methods get a JNIEnv *, not a direct JNIEnv)

given JNIEnv x, you can write ((void**)x)[N] to access a function pointer, where N is:

     4	GetVersion
     5	DefineClass
 6	FindClass

	import os, stat
	from contextlib import contextmanager
	from subprocess import run

	@contextmanager
	def fd_context(fd: int):
	try:
	yield fd
	finally:
	os.close(fd)

	open import Level using (suc; _⊔_)
	open import Function using (id; _∘_)

	open import Data.List as List using (
	List; []; _∷_; [_]; map; reverse; align; alignWith; foldr; head; last; drop; length
	)
	import Data.List.Properties as List
	import Data.List.Relation.Unary.All as All
	import Data.List.Relation.Binary.Pointwise as PW

	import java.awt.*;
	import java.awt.datatransfer.*;
	import java.awt.image.*;
	import java.io.*;

	public class Main {
	public static void main(String[] arg) {
	var flavorMap = (FlavorTable) SystemFlavorMap.getDefaultFlavorMap();
	System.out.println("formats for image flavor: ");
	for (var format : flavorMap.getNativesForFlavor(DataFlavor.imageFlavor))

	use std::{ops::{Add, AddAssign}, convert::TryInto};

	/// In-place unstable radix sort with sentinel
	///
	/// - `L` is the type used for bin counts (must be able to hold `arr.len()`).
	/// - `key: F` is a function taking an array item + a level, and returning an alphabet symbol.
	/// - `AL` is the alphabet length, i.e. number of bins. it must hold that `key(...) < AL`.
	///
	/// 0 is the sentinel, meaning that for any item, `key(item, max_level) == 0`.
	/// `max_level` may differ among items.

	'''
	Reads the contents of a TCP stream carrying [one side of] an
	RTMP connection, and blindly dumps streams above 0 in an FLV.

	$ ./rtmp2flv.py < tcp-stream > video.flv
	'''

	from enum import IntEnum, unique
	from dataclasses import dataclass, field
	from typing import NamedTuple, Self, Optional

	// compile with: clang -Wpedantic -Wall -Wextra $(pkg-config --cflags libpcap) sllv2tosll.c $(pkg-config --libs libpcap) -o sllv2tossl

	#include <errno.h>
	#include <string.h>
	#include <stdio.h>
	#include <stdint.h>
	#include <stdlib.h>
	#include <pcap/pcap.h>

	char error_buffer[PCAP_ERRBUF_SIZE];

	{-# LANGUAGE TupleSections #-}
	import Data.Array (accumArray, listArray, elems)
	import Data.List.Extra (findIndices, groupSortOn, chunksOf)
	import Data.Vector ((!), toList)
	import Control.Monad (forM_, when, zipWithM)
	import qualified Data.Vector as Vec
	import qualified Data.Vector.Mutable as MVec

	fill v = zipWithM (MVec.write v) [0..MVec.length v - 1]

	use std::{mem, ptr};
	use crate::BufferAlloc;

	unsafe fn write_ptr(n: mut usize, ptr: Option<mut usize>) {
	debug_assert!(!ptr.is_some_and(\|ptr\| n == ptr));
	ptr::write(n.cast(), ptr.unwrap_or(n))
	}

	unsafe fn read_ptr(n: mut usize) -> Option<mut usize> {
	let ptr = ptr::read(n.cast());