seqpacker 0.1.2

The core of SeqPacker, written in Rust. This crate provides 11 bin-packing algorithms for packing variable-length sequences into fixed-size bins, reducing padding waste from 20-40% down to 1-5%.

Python bindings are available via the seqpacker PyPI package.

Quick Start

use seqpacker::{Packer, PackStrategy, Sequence};

let packer = Packer::new(2048)
    .with_strategy(PackStrategy::OptimizedBestFitDecreasing);

let sequences = vec![
    Sequence::new(0, 500),
    Sequence::new(1, 600),
    Sequence::new(2, 400),
    Sequence::new(3, 1000),
];

let result = packer.pack(sequences).unwrap();
println!("Bins: {}", result.packs.len());
println!("Efficiency: {:.2}%", result.metrics.efficiency * 100.0);

Pack from lengths

use seqpacker::{Packer, PackStrategy};

let packer = Packer::new(1024)
    .with_strategy(PackStrategy::FirstFitDecreasing);

let result = packer.pack_lengths(&[1000, 800, 600, 500, 400, 300, 200, 100]).unwrap();

for (i, pack) in result.packs.iter().enumerate() {
    let ids: Vec<usize> = pack.sequences.iter().map(|s| s.id).collect();
    println!("Pack {}: items {:?}, used {}/{}", i, ids, pack.used_capacity(), pack.capacity);
}

Streaming

For online / bounded-space packing, use StreamPacker with NextFit or Harmonic:

use seqpacker::{StreamPacker, StreamStrategy};

let mut stream = StreamPacker::new(2048, StreamStrategy::NextFit);

let lengths = [500, 600, 1500, 400];

for &len in &lengths {
    for completed_pack in stream.add(len).unwrap() {
        println!("Completed: {} items, {}/{} used", completed_pack.len(), completed_pack.used_capacity(), completed_pack.capacity);
    }
}

// Flush remaining
for pack in stream.finish() {
    println!("Remaining: {} items", pack.len());
}

Algorithms

11 bin-packing algorithms from O(n) online to near-optimal offline:

Select an algorithm via PackStrategy:

use seqpacker::{Packer, PackStrategy};

let packer = Packer::new(2048)
    .with_strategy(PackStrategy::ModifiedFirstFitDecreasing);

Performance

SeqPacker achieves equal packing efficiency to competitors while being significantly faster:

See the interactive benchmark dashboard for detailed results.

License

MIT