Benchmark plural

plural benchmarks

In this benchmark we’ll look at two metrics : Library size and speed.

When doing the other benchmark in this repository I saw that what was taking most of the time wasn’t the parsing or rendering of messageformat but the different formatters and plurals generators.

I was curious about how different they were.

Libraries size

Sources can be found in src, measure taken on 25/06/2020 With latest available versions

Npm Package Version Size Comment
@onigoetz/intl-formatters 1.0.0-rc.2 2.6K Embeds the formatters in the runtime
@onigoetz/make-plural 1.0.0-rc.2 3.6K Contains only the formatter for en in this example, have to be shipped separately.
make-plural 7.4.0 16K Contains all locales
@phensley/plurals 1.9.0 33K Contains all locales

Benchmarks run on

  • Node.js v20.9.0
  • Apple M2 CPU
  • October 17, 2024

Ordinal

const input = [`ordinal`, `en`, 2];

// Renders: `two`
Name ops/sec MoE Runs sampled
make-plural 14,325,529 ± 0.55% 99
@phensley/plurals 13,854,078 ± 0.62% 95
@onigoetz/make-plural(memo) 11,552,355 ± 0.76% 95
@onigoetz/intl 2,131,691 ± 0.81% 94
@onigoetz/make-plural 54,598 ± 2.94% 98

Choosing a library for your use case

  • If you know which languages you need in advance: @phensley/plurals and make-plural both ship a pre-compiled and pre-optimized set of rules for plurals. They’re the fastest options and will be the smallest if your build system can perform Tree Shaking.
  • If you do not know the languages in advance.
    • And speed is a concern: @onigoetz/make-plural and its memoized alternative @onigoetz/make-plural(memo) should do the trick. you will need to ship the pluralization rules one way or another to the function.
    • And speed is not a concern: @onigoetz/intl uses Intl.PluralRules that ships with most browsers and Node.js runtimes but as you can see it’s much slower than other solutions.