nordisk/znippy-zoomies

lgz/examples/bitcheck.rs

//! Debug aid: are the split starts real block starts? `bitcheck <file.gz> <n>`
fn main() {
    let a: Vec<String> = std::env::args().collect();
    let gz = std::fs::read(&a[1]).unwrap();
    let n: usize = a[2].parse().unwrap();
    let s = lgz::parallel::deflate_offset(&gz).unwrap();
    let d = &gz[s..gz.len() - 8];
    let isz = u32::from_le_bytes(gz[gz.len() - 4..].try_into().unwrap()) as usize;
    // every true block start, by decoding block by block
    let mut out = vec![0u8; isz + linflate::OVERWRITE_HEADROOM];
    let mut starts = std::collections::BTreeMap::new();
    let (mut bit, mut done) = (0usize, 0usize);
    loop {
        let h = (u16::from_le_bytes([d[bit / 8], d[bit / 8 + 1]]) >> (bit % 8)) & 7;
        starts.insert(bit, h);
        let win = done.min(32768);
        let mut buf = vec![0u8; win + isz - done + linflate::OVERWRITE_HEADROOM];
        buf[..win].copy_from_slice(&out[done - win..done]);
        let r = linflate::inflate_bits(d, bit, bit + 1, &mut buf, win).unwrap();
        out[done..done + r.written].copy_from_slice(&buf[win..win + r.written]);
        done += r.written;
        if r.final_block {
            break;
        }
        bit = r.end_bit;
    }
    let types: Vec<u16> = starts.values().copied().collect();
    println!(
        "blocks {} (dynamic {}, fixed {}, stored {})",
        types.len(),
        types.iter().filter(|&&h| h >> 1 == 2).count(),
        types.iter().filter(|&&h| h >> 1 == 1).count(),
        types.iter().filter(|&&h| h >> 1 == 0).count()
    );
    let sp = lgz::bitsplit::split_chunk_bits(d, 0, n);
    for &b in &sp.starts {
        let real = starts.contains_key(&b);
        let next = starts.range(b..).next().map(|(k, _)| *k);
        println!("start {b} real={real} next_true={next:?}");
    }
}