Skip to content

Advanced Configuration Guide

This guide explains how to customize kand for numerical precision, optimization, and cross-platform use. Custom configurations require building from source with maturin, as the default pip install kand provides a pre-built package with fixed settings (f64, i64, check).

Customization Options

Numerical Precision

Choose the floating-point precision for your needs:

  • f32 (32-bit): Lower memory usage, great for large datasets or constrained systems.
  • f64 (64-bit): Default, higher precision for complex calculations.
Precision Memory Usage Precision Use Case
f32 Low Lower Embedded systems, big data
f64 Medium Higher Scientific computing

Integer Types

Select the integer type for indexing:

  • i32 (32-bit): Memory-efficient, suits smaller applications.
  • i64 (64-bit): Default, handles larger datasets.
Type Description Best For
i32 32-bit integers Small-scale applications
i64 64-bit integers Large datasets

Validation Levels

Adjust validation for safety vs. performance:

  • check: Basic validation, ideal for production.
  • deep-check: Detailed checks, best for debugging.
  • None: No validation, fastest but risky.
Level Safety Performance Use Case
check High Medium Production
deep-check Highest Slowest Debugging
None Low Fastest Tested environments

Performance vs. Safety

Skipping validation boosts speed but may cause issues with invalid inputs. Use cautiously.

Building from Source

Customizing kand requires a local build with maturin.

Prerequisites

  • Python: 3.8+
  • Rust: 1.80+
  • maturin: pip install maturin

Build Steps

  1. Clone the Repository: 

    git clone https://github.com/rust-ta/kand.git
cd kand

  2. Build with Custom Features: For development (editable install): 

    maturin develop --features f32,i64,check

High-Performance Build Example

Use --release for an optimized build: 

maturin build --release --features f64,i64,check

Troubleshooting

  • Feature Not Working: Ensure --features matches your desired configuration (e.g., f32,i64,check).
  • Build Fails: Check Rust/Python versions or run cargo build to debug.
  • Help: Visit GitHub Discussions.

Performance Trade-offs

Configuration Memory Speed Precision
f32, i32, no checks Lowest Fastest Lowest
f32, i64, check Low Fast Low
f64, i64, check Medium Medium High
f64, i64, deep-check High Slowest Highest

See the Performance Guide for benchmarks.