use std::{slice::ChunksExact, sync::OnceLock}; const ALPHABET: &[u8] = b"0123456789=ABCDEFGHIJKLMNOPQRSTUVWXYZ_abcdefghijklmnopqrstuvwxyz"; pub fn gwilym_decode(input: &str) -> anyhow::Result> { GwilymDecodeIter::new(input) } pub struct GwilymDecodeIter<'a> { chunks: ChunksExact<'a, u8>, } impl<'a> GwilymDecodeIter<'a> { fn new(input: &'a str) -> anyhow::Result { let input = input .strip_prefix("https://agbrs.dev/crash#") .unwrap_or(input); let Some((input, version)) = input.rsplit_once('v') else { anyhow::bail!("Does not contain version"); }; if version != "1" { anyhow::bail!("Only version 1 is supported"); } if input.len() % 3 != 0 { anyhow::bail!("Input string must have length a multiple of 3"); } Ok(Self { chunks: input.as_bytes().chunks_exact(3), }) } } impl<'a> Iterator for GwilymDecodeIter<'a> { type Item = u32; fn next(&mut self) -> Option { let Some(chunk) = self.chunks.next() else { return None; }; let value = decode_chunk(chunk); if value & (1 << 16) != 0 { let upper_bits = value << 16; let lower_bits = self.next().unwrap_or(0) & 0xffff; return Some(upper_bits | lower_bits); } Some(value | 0x0800_0000) } } fn decode_chunk(chunk: &[u8]) -> u32 { let a = get_value_for_char(chunk[0]); let b = get_value_for_char(chunk[1]); let c = get_value_for_char(chunk[2]); (a << (16 - 5)) | (b << (16 - 10)) | c } fn get_value_for_char(input: u8) -> u32 { static REVERSE_ALHPABET: OnceLock<[u8; 128]> = OnceLock::new(); REVERSE_ALHPABET.get_or_init(|| { let mut result = [0; 128]; for (i, &c) in ALPHABET.iter().enumerate() { result[c as usize] = i as u8; } result })[input as usize] as u32 } #[cfg(test)] mod test { use super::{gwilym_decode, ALPHABET}; use std::fmt::Write; #[test] fn should_correctly_decode_16s() -> anyhow::Result<()> { assert_eq!( &gwilym_decode("2QI65Q69306Kv1")?.collect::>(), &[0x0800_16d3, 0x0800_315b, 0x0800_3243, 0x0800_0195] ); Ok(()) } fn encode_16(input: u16) -> [u8; 3] { let input = input as usize; [ ALPHABET[input >> (16 - 5)], ALPHABET[(input >> (16 - 10)) & 0b11111], ALPHABET[input & 0b111111], ] } fn encode_32(input: u32) -> [u8; 6] { let input = input as usize; let output_lower_16 = encode_16(input as u16); let input_upper_16 = input >> 16; [ ALPHABET[(input_upper_16 >> (16 - 5)) | (1 << 5)], ALPHABET[(input_upper_16 >> (16 - 10)) & 0b11111], ALPHABET[input_upper_16 & 0b111111], output_lower_16[0], output_lower_16[1], output_lower_16[2], ] } #[test] fn should_correctly_decode_16s_and_32s() -> anyhow::Result<()> { let trace: &[u32] = &[ 0x0300_2990, 0x0800_3289, 0x0500_2993, 0x3829_2910, 0xffff_ffff, 0x0000_0000, ]; let mut result = String::new(); for &ip in trace { if ip & 0xFFFF_0000 == 0x0800_0000 { let encoded = encode_16(ip as u16); let encoded_s = std::str::from_utf8(&encoded)?; write!(&mut result, "{encoded_s}")? } else { let encoded = encode_32(ip); let encoded_s = std::str::from_utf8(&encoded)?; write!(&mut result, "{encoded_s}")? } } write!(&mut result, "v1")?; assert_eq!(&gwilym_decode(&result)?.collect::>(), trace); Ok(()) } #[test] fn should_strip_the_agbrsdev_prefix() -> anyhow::Result<()> { assert_eq!( &gwilym_decode("https://agbrs.dev/crash#2QI65Q69306Kv1")?.collect::>(), &[0x0800_16d3, 0x0800_315b, 0x0800_3243, 0x0800_0195] ); Ok(()) } }