NumPack

NumPack is a lightning-fast array manipulation engine that revolutionizes how you handle large-scale NumPy arrays. By combining Rust's raw performance with Python's ease of use, NumPack delivers up to 20x faster operations than traditional methods, while using minimal memory. Whether you're working with gigabyte-sized matrices or performing millions of array operations, NumPack makes it effortless with its zero-copy architecture and intelligent memory management.

Key highlights:

• 🚀 Up to 20x faster than traditional NumPy storage methods
• 💾 Zero-copy operations for minimal memory footprint
• 🔄 Seamless integration with existing NumPy workflows
• 🛠 Battle-tested in production with arrays exceeding 1 billion rows

Features

• High Performance: Optimized for both reading and writing large numerical arrays
• Memory Mapping Support: Efficient memory usage through memory mapping capabilities
• Selective Loading: Load only the arrays you need, when you need them
• In-place Operations: Support for in-place array modifications without full file rewrite
• Parallel I/O: Utilizes parallel processing for improved performance
• Multiple Data Types: Supports various numerical data types including:
  • Boolean
  • Unsigned integers (8-bit to 64-bit)
  • Signed integers (8-bit to 64-bit)
  • Floating point (32-bit and 64-bit)

Installation

From PyPI (Recommended)

Prerequisites

• Python >= 3.9
• NumPy >= 1.26.0

pip install numpack

From Source

To build and install NumPack from source, you need to meet the following requirements:

Prerequisites

• Python >= 3.9
• Rust >= 1.70.0
• NumPy >= 1.26.0
• Appropriate C/C++ compiler (depending on your operating system)
  • Linux: GCC or Clang
  • macOS: Clang (via Xcode Command Line Tools)
  • Windows: MSVC (via Visual Studio or Build Tools)

Build Steps

1. Clone the repository:

git clone https://github.com/BirchKwok/NumPack.git
cd NumPack

2. Install maturin (for building Rust and Python hybrid projects):

pip install maturin>=1.0,<2.0

3. Build and install:

# Install in development mode
maturin develop

# Or build wheel package
maturin build --release
pip install target/wheels/numpack-*.whl

Platform-Specific Notes

• Linux Users:

  • Ensure python3-dev (Ubuntu/Debian) or python3-devel (Fedora/RHEL) is installed
  • If using conda environment, make sure the appropriate compiler toolchain is installed

• macOS Users:

  • Make sure Xcode Command Line Tools are installed: xcode-select --install
  • Supports both Intel and Apple Silicon architectures

• Windows Users:

  • Visual Studio or Visual Studio Build Tools required
  • Ensure "Desktop development with C++" workload is installed

Usage

Basic Operations

import numpy as np
from numpack import NumPack

# Create a NumPack instance
npk = NumPack("data_directory")

# Save arrays
arrays = {
    'array1': np.random.rand(1000, 100).astype(np.float32),
    'array2': np.random.rand(500, 200).astype(np.float32)
}
npk.save(arrays)

# Load arrays
# Normal mode
loaded = npk.load("array1")

# Memory mapping mode for large arrays
with npk.mmap_mode() as mmap_npk:
   # Access specific arrays
   array1 = mmap_npk.load('array1')
   array2 = mmap_npk.load('array2')

Advanced Operations

# Replace specific rows
replacement = np.random.rand(10, 100).astype(np.float32)
npk.replace({'array1': replacement}, [0, 1, 2, 3, 4, 5, 6, 7, 8, 9])  # Using list indices
npk.replace({'array1': replacement}, slice(0, 10))  # Using slice notation

# Append new arrays
new_arrays = {
    'array3': np.random.rand(200, 100).astype(np.float32)
}
npk.append(new_arrays)

# Drop arrays or specific rows
npk.drop('array1')  # Drop entire array
npk.drop(['array1', 'array2'])  # Drop multiple arrays
npk.drop('array2', [0, 1, 2])  # Drop specific rows

# Random access operations
data = npk.getitem('array1', [0, 1, 2])  # Access specific rows
data = npk.getitem('array1', slice(0, 10))  # Access using slice
data = npk['array1']  # Dictionary-style access for entire array

# Metadata operations
shapes = npk.get_shape()  # Get shapes of all arrays
shapes = npk.get_shape('array1')  # Get shape of specific array
members = npk.get_member_list()  # Get list of array names
mtime = npk.get_modify_time('array1')  # Get modification time
metadata = npk.get_metadata()  # Get complete metadata

# Stream loading for large arrays
for batch in npk.stream_load('array1', buffer_size=1000):
    # Process 1000 rows at a time
    process_batch(batch)

# Reset/clear storage
npk.reset()  # Clear all arrays

# Iterate over all arrays
for array_name in npk:
    data = npk[array_name]
    print(f"{array_name} shape: {data.shape}")

Lazy Loading and Buffer Operations

NumPack supports lazy loading and buffer operations, which are particularly useful for handling large-scale datasets. Using the lazy=True parameter enables data to be loaded only when actually needed, making it ideal for streaming processing or scenarios where only partial data access is required.

from numpack import NumPack
import numpy as np

# Create NumPack instance and save large-scale data
npk = NumPack("test_data/", drop_if_exists=True)
a = np.random.random((1000000, 128))  # Create a large array
npk.save({"arr1": a})

# Lazy loading - keeps data in buffer
lazy_array = npk.load("arr1", lazy=True)  # LazyArray Object

# Perform computations with lazy-loaded data
# Only required data is loaded into memory
similarity_scores = np.inner(a[0], npk.load("arr1", lazy=True))

Memory Mapping Mode

For large arrays, memory mapping mode provides more efficient memory usage:

# Using memory mapping mode
with npk.mmap_mode() as mmap_npk:
    # Access specific arrays
    array1 = mmap_npk.load('array1')  # Array is not fully loaded into memory
    array2 = mmap_npk.load('array2')
    
    # Perform operations on memory-mapped arrays
    result = array1[0:1000] + array2[0:1000]

Performance

NumPack offers significant performance improvements compared to traditional NumPy storage methods, especially in data modification operations and random access. Below are detailed benchmark results:

Benchmark Results

The following benchmarks were performed on an MacBook Pro (Apple Silicon) with arrays of size 1M x 10 and 500K x 5 (float32).

Storage Operations

Operation	NumPack	NumPy NPZ	NumPy NPY
Save	0.014s (0.86x NPZ, 0.57x NPY)	0.012s	0.008s
Full Load	0.008s (1.63x NPZ, 0.88x NPY)	0.013s	0.007s
Selective Load	0.006s (1.50x NPZ, -)	0.009s	-
Mmap Load	0.006s (2.00x NPZ, 0.67x NPY)	0.012s	0.004s

Data Modification Operations

Operation	NumPack	NumPy NPZ	NumPy NPY
Single Row Replace	0.000s (19.00x NPZ, 12.00x NPY)	0.019s	0.012s
Continuous Rows (10K)	0.001s (20.00x NPZ, 12.00x NPY)	0.020s	0.012s
Random Rows (10K)	0.015s (1.33x NPZ, 0.87x NPY)	0.020s	0.013s
Large Data Replace (500K)	0.019s (1.00x NPZ, 0.74x NPY)	0.019s	0.014s

Drop Operations

Operation (1M rows, float32)	NumPack	NumPy NPZ	NumPy NPY
Drop Array	0.001s (22.00x NPZ, 1.00x NPY)	0.022s	0.001s
Drop First Row	0.014s (3.21x NPZ, 1.93x NPY)	0.045s	0.027s
Drop Last Row	0.000s (∞x NPZ, ∞x NPY)	0.045s	0.027s
Drop Middle Row	0.014s (3.21x NPZ, 1.93x NPY)	0.045s	0.027s
Drop Front Continuous (10K rows)	0.016s (2.81x NPZ, 1.69x NPY)	0.045s	0.027s
Drop Middle Continuous (10K rows)	0.016s (2.81x NPZ, 1.69x NPY)	0.045s	0.027s
Drop End Continuous (10K rows)	0.001s (45.00x NPZ, 27.00x NPY)	0.045s	0.027s
Drop Random Rows (10K rows)	0.018s (2.50x NPZ, 1.50x NPY)	0.045s	0.027s
Drop Near Non-continuous (10K rows)	0.015s (3.00x NPZ, 1.80x NPY)	0.045s	0.027s

Append Operations

Operation	NumPack	NumPy NPZ	NumPy NPY
Small Append (1K rows)	0.000s (22.00x NPZ, 18.00x NPY)	0.022s	0.018s
Large Append (500K rows)	0.003s (9.67x NPZ, 6.67x NPY)	0.029s	0.020s

Random Access Performance (10K indices)

Operation	NumPack	NumPy NPZ	NumPy NPY
Random Access	0.008s (2.00x NPZ, 1.38x NPY)	0.016s	0.011s

File Size Comparison

Format	Size	Ratio
NumPack	47.68 MB	1.0x
NPZ	47.68 MB	1.0x
NPY	47.68 MB	1.0x

Large-scale Data Operations (>1B rows, Float32)

Operation	NumPack	NumPy NPZ	NumPy NPY
Replace	Zero-copy in-place modification	Memory exceeded	Memory exceeded
Drop	Zero-copy in-place deletion	Memory exceeded	Memory exceeded
Append	Zero-copy in-place addition	Memory exceeded	Memory exceeded
Random Access	Near-hardware I/O speed	Memory exceeded	Memory exceeded

Matrix Computation Performance (1M rows x 128 columns, Float32)

Operation	NumPack	NumPy NPZ	NumPy NPY	In-Memory
Inner Product	0.019s (6.58x NPZ, 1.00x NPY)	0.125s	0.019s	0.011s
Other calculations are similar to the above case	...	...	...	...

Key Advantage: NumPack achieves the same performance as NumPy's NPY mmap (0.019s) for matrix computations, with several implementation advantages:

• Uses Arc for reference counting, ensuring automatic resource cleanup
• Implements MMAP_CACHE to avoid redundant data loading
• Linux-specific optimizations with huge pages and sequential access hints
• Supports parallel I/O operations for improved data throughput
• Optimizes memory usage through Buffer Pool to reduce fragmentation

Key Performance Highlights

1. Data Modification:

   • Single row replacement: NumPack is 19x faster than NPZ and 12x faster than NPY
   • Continuous rows: NumPack is 20x faster than NPZ and 12x faster than NPY
   • Random rows: NumPack is 1.33x faster than NPZ but 0.87x slower than NPY
   • Large data replacement: NumPack is comparable to NPZ but 0.74x slower than NPY

2. Drop Operations:

   • Drop array: NumPack is 22x faster than NPZ and comparable to NPY
   • Drop rows: NumPack is currently 0.61x slower than NPZ and 0.41x slower than NPY
   • NumPack provides efficient in-place row deletion without full file rewrite

3. Append Operations:

   • Small append (1K rows): NumPack is 22x faster than NPZ and 18x faster than NPY
   • Large append (500K rows): NumPack is 9.67x faster than NPZ and 6.67x faster than NPY
   • NumPack excels at both small and large append operations

4. Loading Performance:

   • Full load: NumPack is 1.63x faster than NPZ but 0.88x slower than NPY
   • Memory-mapped load: NumPack is 2.00x faster than NPZ but 0.67x slower than NPY
   • Selective load: NumPack is 1.50x faster than NPZ

5. Random Access:

   • NumPack is 2.00x faster than NPZ and 1.38x faster than NPY for random index access

6. Storage Efficiency:

   • All formats achieve identical compression ratios (47.68 MB)
   • NumPack maintains high performance while keeping file sizes competitive

7. Matrix Computation:

   • NumPack matches NPY mmap performance while providing better resource management
   • 6.58x faster than NPZ mmap for matrix operations
   • Only 1.72x slower than pure in-memory computation
   • Zero risk of file descriptor leaks or resource exhaustion

Note: All benchmarks were performed with float32 arrays. Performance may vary depending on data types, array sizes, and system configurations. Numbers greater than 1.0x indicate faster performance, while numbers less than 1.0x indicate slower performance.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the Apache License, Version 2.0 - see the LICENSE file for details.

Copyright 2024 NumPack Contributors

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.