buffered read vs non buffered

[Its-Just-Nans]

Hello

Is this normal that when we use. A buffer, we don't have the same result as without using it ?

When using data.len() > buffer_size (8k), the buffer need one more read to completely read the file

Is this normal?

use flate2::bufread::ZlibDecoder;
use flate2::write::ZlibEncoder;
use flate2::{Compression, Decompress};
use std::io::prelude::{Read, Write};

fn main() {
    let mut e = ZlibEncoder::new(Vec::new(), Compression::new(6));
    // CONTENT should be superior of the buffer size
    let mut data = vec![0; 1024 * 8];
    getrandom::fill(&mut data).unwrap();
    let writen = e.write(&data).unwrap();
    println!("writen {writen}");
    let compressed_bytes = e.finish().unwrap();
    println!("compressed size = {}", compressed_bytes.len());

    println!("Test with header");
    let mut d = ZlibDecoder::new(&*compressed_bytes);
    let mut disk_buffer = [0; 1024];
    loop {
        let res = d.read(&mut disk_buffer[..]).unwrap();
        println!("count read = {res}");
        if res == 0 {
            break;
        }
    }
    println!("Test with header and buffered");
    let mut d = ZlibDecoder::new(std::io::BufReader::new(&*compressed_bytes));
    let mut disk_buffer = [0; 1024];
    loop {
        let res = d.read(&mut disk_buffer[..]).unwrap();
        println!("count read = {res}");
        if res == 0 {
            break;
        }
    }

    let end = compressed_bytes.len() - 4;
    let raw_buf = &compressed_bytes[2..end];
    println!("raw_buf = {}", raw_buf.len());

    println!("Test with raw access");
    let mut d = ZlibDecoder::new_with_decompress(raw_buf, Decompress::new(false));
    let mut disk_buffer = [0; 1024];
    loop {
        let res = d.read(&mut disk_buffer[..]).unwrap();
        println!("count read = {res}");
        if res == 0 {
            break;
        }
    }

    println!("Test with buffered access");
    let mut d =
        ZlibDecoder::new_with_decompress(std::io::BufReader::new(raw_buf), Decompress::new(false));
    let mut disk_buffer = [0; 1024];
    loop {
        let res = d.read(&mut disk_buffer[..]).unwrap();
        println!("count read = {res}");
        if res == 0 {
            break;
        }
    }
}

writen 8192
compressed size = 8203
Test with header
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 0
Test with header and buffered
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1017
count read = 7
count read = 0
raw_buf = 8197
Test with raw access
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 0
Test with buffered access
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1024
count read = 1019
count read = 5
count read = 0

Thanks

[Its-Just-Nans]

It seems that it's the normal behavior of read() and if we want to read the full buffer we need read_exact()

https://doc.rust-lang.org/stable/std/io/trait.Read.html#tymethod.read