Mustache vs. Liquid templates benchmark (naïve)

01_mustache_liquid_benchmark.rb

# This is a very basic benchmark of Mustache (1.0.2) vs.
# Liquid (3.0.6) rendering. It does two sets of benchmarks: One with
# "precompiled" templates (i.e. the Mustache::Template or
# Liquid::Template object is kept between runs) and one without (a new
# Template object is instantiated for each run). This benchmark tests
# a basic loop; no other features e.g. partials or conditionals
# are tested.

require "benchmark/ips"
require "mustache"
require "liquid"

mustache_template = <<END
{{#products}}
  <div class='product_brick'>
    <div class='container'>
      <div class='element'>
        <img src='images/{{image}}' class='product_miniature' />
      </div>
      <div class='element description'>
        <a href={{url}} class='product_name block bold'>
          {{external_index}}
        </a>
      </div>
    </div>
  </div>
{{/products}}
END

liquid_template = <<END
{% for product in products %}
  <div class='product_brick'>
    <div class='container'>
      <div class='element'>
        <img src='images/{{ product.image }}' class='product_miniature' />
      </div>
      <div class='element description'>
        <a href={{ product.url }} class='product_name block bold'>
          {{ product.external_index }}
        </a>
      </div>
    </div>
  </div>
{% endfor %}
END

data = {
  'products' => 100.times.map do |n|
    { 'image' => "foo-#{rand 100}.png", 'url' => "http://bar-#{rand 100}.com", 'external_index' => "FOO #{rand 100}" }
  end
}

def render_mustache(template, data)
  tmpl = Mustache::Template.new(template)
  ctx = Mustache::Context.new(Mustache.new)
  ctx.push(data)
  tmpl.render(ctx)
end

def render_liquid(template, data)
  Liquid::Template.parse(template).render(data)
end

mustache_template_pre = Mustache::Template.new(mustache_template)
liquid_template_pre = Liquid::Template.parse(liquid_template)

def render_mustache_pre(tmpl, data)
  ctx = Mustache::Context.new(Mustache.new)
  ctx.push(data)
  tmpl.render(ctx)
end

def render_liquid_pre(tmpl, data)
  tmpl.render(data)
end

Benchmark.ips do |x|
  x.report('render_mustache') { render_mustache(mustache_template, data) }
  x.report('render_mustache_pre') { render_mustache_pre(mustache_template_pre, data) }
  x.report('render_liquid') { render_liquid(liquid_template, data) }
  x.report('render_liquid_pre') { render_liquid_pre(liquid_template_pre, data) }
  x.compare!
end

02_result.txt

Warming up --------------------------------------
     render_mustache    33.000  i/100ms
 render_mustache_pre    49.000  i/100ms
       render_liquid    24.000  i/100ms
   render_liquid_pre    25.000  i/100ms
Calculating -------------------------------------
     render_mustache    330.242  (± 7.3%) i/s -      1.650k
 render_mustache_pre    505.154  (± 6.1%) i/s -      2.548k
       render_liquid    240.000  (± 7.5%) i/s -      1.200k
   render_liquid_pre    248.057  (± 6.0%) i/s -      1.250k

Comparison:
 render_mustache_pre:      505.2 i/s
     render_mustache:      330.2 i/s - 1.53x slower
   render_liquid_pre:      248.1 i/s - 2.04x slower
       render_liquid:      240.0 i/s - 2.10x slower

Running this Liquid-C is much more faster:

# add these line 
require 'liquid'
require 'liquid/c'

New benchmark:

# mustache (1.1.1)
# liquid (5.1.0)
# liquid-c (4.1.0)
Warming up --------------------------------------
     render_mustache    95.000  i/100ms
 render_mustache_pre   155.000  i/100ms
       render_liquid   368.000  i/100ms
   render_liquid_pre   410.000  i/100ms
Calculating -------------------------------------
     render_mustache    948.676  (± 0.5%) i/s -      4.750k in   5.007094s
 render_mustache_pre      1.554k (± 0.4%) i/s -      7.905k in   5.087068s
       render_liquid      3.695k (± 1.3%) i/s -     18.768k in   5.080867s
   render_liquid_pre      4.150k (± 1.3%) i/s -     20.910k in   5.039370s

Comparison:
   render_liquid_pre:     4150.1 i/s
       render_liquid:     3694.6 i/s - 1.12x  (± 0.00) slower
 render_mustache_pre:     1554.0 i/s - 2.67x  (± 0.00) slower
     render_mustache:      948.7 i/s - 4.37x  (± 0.00) slower