Skip to content

File iterator or stream with auto or manual chunk size selection

License

Notifications You must be signed in to change notification settings

m62624/get_chunk

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

get_chunk

Crates.io Static Badge docs.rs Codecov

About

get_chunk is a library for creating file iterators or streams (asynchronous iterators), specialized in efficient file chunking. The main task, the ability to retrieve chunk data especially from large files.

Key Features:

  • File Chunking: Divide files, including large ones, into seamless chunks with each "Next" iteration.
  • Modes: Choose between automatic tuning or manually set chunk size based on percentage or byte count.
  • Automatic chunking: Each "Next" iteration dynamically determines an optimal chunk size, facilitating efficient handling of even large files.

⚠️ Important Notice:

The algorithm adjusts chunk sizes for optimal performance after the "Next" call, taking into account available RAM. However, crucially, this adjustment occurs only after the current chunk is sent and before the subsequent "Next" call.


It's important to note a potential scenario: Suppose a chunk is 15GB, and there's initially 16GB of free RAM. If, between the current and next "Next" calls, 2GB of RAM becomes unexpectedly occupied, the current 15GB chunk will still be processed. This situation introduces a risk, as the system might either reclaim resources (resulting in io::error) or lead to a code crash.


Iterators created by get_chunk don't store the entire file in memory, especially for large datasets. Their purpose is to fetch data from files in chunks, maintaining efficiency.

Key Points:

  • Limited File Retention: Creating an iterator for a small file might result in fetching all data, OS-dependent. However, this doesn't guarantee file persistence after iterator creation.
  • Deletion Warning: Deleting a file during iterator or stream iterations will result in an error. These structures don't track the last successful position.
  • No File Restoration: Attempting to restore a deleted file during iterations is not supported. These structures don't keep track of the file's original state.

Iterator version


Example

use get_chunk::iterator::FileIter;
// Note: requires a `size_format` attribute.
use get_chunk::data_size_format::iec::IECUnit;

fn main() -> std::io::Result<()> {

    let file_iter = FileIter::new("file.txt")?;
    // or 
    // let file_iter = FileIter::try_from(File::open("file.txt")?)?;
    // ...
    for chunk in file_iter {
        match chunk {
            Ok(data) => {
              // some calculations with chunk
              //.....
              println!("{}", IECUnit::auto(data.len() as f64));
            }
            Err(_) => break,
        }
    }
    
    Ok(())
}

Example

use get_chunk::{
  // Note: requires a `size_format` attribute.
    data_size_format::iec::{IECSize, IECUnit},
    iterator::FileIter,
    ChunkSize,
};

fn main() -> std::io::Result<()> {
    let file_iter = FileIter::new("file.txt")?
        .include_available_swap()
        .set_mode(ChunkSize::Bytes(40000))
        .set_start_position_bytes(IECUnit::new(432.0, IECSize::Mebibyte).into());
    Ok(())
}

Stream version

Example

// Note: requires the `size_format` and `stream` attributes.
use get_chunk::data_size_format::iec::IECUnit;
use get_chunk::stream::{FileStream, StreamExt};

#[tokio::main]
async fn main() -> std::io::Result<()> {

    let mut file_stream = FileStream::new("file.txt").await?;
    // or
    // let mut file_stream = FileStream::try_from_data(File::open("file.txt").await?)?;
    // ...
    while let Ok(chunk) = file_stream.try_next().await {
        match chunk {
            Some(chunk) => {
                 // some calculations with chunk
            },
            None => break,
        }
    }

    Ok(())
}

How it works

The calculate_chunk function in the ChunkSize enum determines the optimal chunk size based on various parameters. Here's a breakdown of how the size is calculated:

The variables prev and now represent the previous and current read time, respectively.

prev:

Definition: prev represents the time taken to read a piece of data in the previous iteration.

now:

Definition: now represents the current time taken to read the data fragment in the current iteration.

  1. Auto Mode:

    • If the previous read time (prev) is greater than zero:
      • If the current read time (now) is also greater than zero:
        • If now is less than prev, decrease the chunk size using decrease_chunk method.
        • If now is greater than or equal to prev, increase the chunk size using increase_chunk method.
      • If now is zero or negative, maintain the previous chunk size (prev).
    • If the previous read time is zero or negative, use the default chunk size based on the file size and available RAM.
  2. Percent Mode:

    • Calculate the chunk size as a percentage of the total file size using the percentage_chunk method. The percentage is capped between 0.1% and 100%.
  3. Bytes Mode:

    • Calculate the chunk size based on the specified number of bytes using the bytes_chunk method. The size is capped by the file size and available RAM.

Key Formulas:

  • Increase Chunk Size:
(prev * (1.0 + ((now - prev) / prev).min(0.15))).min(ram_available * 0.85).min(f64::MAX)
  • Decrease Chunk Size:
(prev * (1.0 - ((prev - now) / prev).min(0.45))).min(ram_available * 0.85).min(f64::MAX)
  • Default Chunk Size:
(file_size * (0.1 / 100.0)).min(ram_available * 0.85).min(f64::MAX)
  • Percentage Chunk Size:
(file_size * (percentage.min(100.0).max(0.1) / 100.0)).min(ram_available * 0.85)
  • Bytes Chunk Size:
(bytes as f64).min(file_size).min(ram_available * 0.85)

License

MIT License