FastFloatLemire

Eisel-Lemire algorithm for string-to-float conversion in Ruby.

This gem implements the Eisel-Lemire algorithm with additional fast paths for common number formats.

The implementation includes several fast paths that bypass the full Eisel-Lemire algorithm:

1. Small integer fast path - handles "5", "42", "-123" (up to 3 digits)
2. Simple decimal fast path - handles "1.5", "9.99", "199.95" (up to 3+3 digits)
3. Exact power-of-10 fast path - uses precomputed exact powers of 10 (10^0 to 10^22)
4. Removed overhead - no strlen(), single whitespace skip

These optimizations are based on insights from Nigel Tao's Eisel-Lemire blog post.

gem install fast_float_lemire

Or add to your Gemfile:

gem 'fast_float_lemire'

require 'fast_float_lemire'

FastFloatLemire.parse("3.141592653589793")  #=> 3.141592653589793
FastFloatLemire.parse("1.5")                #=> 1.5
FastFloatLemire.parse("1e10")               #=> 10000000000.0

# Bulk parsing
FastFloatLemire.parse_array(["1.0", "2.0", "3.14"])  #=> [1.0, 2.0, 3.14]

Run benchmarks with:

bundle exec ruby benchmarks/comparison.rb

• Eisel-Lemire paper - "Number Parsing at a Gigabyte per Second"
• fast_float C++ library
• Go implementation

MIT