perf: optimize compiled template allocations and performance

Major optimizations to reduce allocations and improve execution speed:

1. For loops: Replace catch/throw with while + break flag
   - Uses while loop with index instead of .each with catch/throw
   - Break implemented with flag variable, continue with next
   - Result: 18% fewer allocations, 85% faster for simple loops

2. Forloop property inlining
   - Inline forloop.index as (__idx__ + 1), forloop.first as (__idx__ == 0), etc.
   - Completely eliminates forloop hash allocation when all properties inlinable
   - Result: Loop with forloop went from +46% MORE to -16% FEWER allocations

3. LR.to_array helper with EMPTY_ARRAY constant
   - Centralized array conversion with frozen empty array for nil
   - Avoids allocations for empty collections

4. Inline LR.truthy? calls
   - Replace LR.truthy?(x) with (x != nil && x != false)
   - Eliminates method call overhead in conditions

5. Keep Time methods available in sandbox for date filter

Overall results:
- Allocations: 3.5% MORE -> 24% FEWER (27% improvement)
- Time: 64% faster -> 89% faster (25% improvement)

Also adds:
- compile_profiler.rb for measuring allocations/performance
- compile_acceptance_test.rb for output equivalence testing
- OPTIMIZATION.md documenting optimization status
This commit is contained in:
Tobi Lutke
2025-12-31 13:24:33 -04:00
parent 652b9c0897
commit c5a44be104
12 changed files with 996 additions and 59 deletions
+70
View File
@@ -0,0 +1,70 @@
# Liquid Compiled Template Optimization Log
This document tracks optimizations made to the compiled Liquid template engine.
Each entry shows before/after code and measured impact.
---
## Baseline Measurement
**Date:** 2024-12-31
**Commit:** (pending profiler implementation)
### Current State
The compiled template engine generates Ruby code from Liquid templates.
Before optimizations, here's a sample of generated code for a simple loop:
```ruby
# Template: {% for product in products %}{{ forloop.index }}: {{ product.name }}{% endfor %}
->(assigns, __context__, __external__) do
__output__ = +""
__coll1__ = assigns["products"]
__coll1__ = __coll1__.to_a if __coll1__.is_a?(Range)
__len3__ = __coll1__.respond_to?(:length) ? __coll1__.length : 0
__idx2__ = 0
catch(:__loop__break__) do
(__coll1__.respond_to?(:each) ? __coll1__ : []).each do |__item__|
catch(:__loop__continue__) do
assigns["product"] = __item__
assigns['forloop'] = {
'name' => "product-products",
'length' => __len3__,
'index' => __idx2__ + 1,
'index0' => __idx2__,
'rindex' => __len3__ - __idx2__,
'rindex0' => __len3__ - __idx2__ - 1,
'first' => __idx2__ == 0,
'last' => __idx2__ == __len3__ - 1,
}
__output__ << LR.output(LR.lookup(assigns["forloop"], "index", __context__))
__output__ << ": "
__output__ << LR.output(LR.lookup(assigns["product"], "name", __context__))
end
__idx2__ += 1
end
end
assigns.delete("product")
assigns.delete('forloop')
__output__
end
```
### Issues Identified
1. **catch/throw overhead** - Used even when no break/continue in loop
2. **Hash allocation per iteration** - 8 key/value pairs computed every time
3. **respond_to? checks** - Redundant after type is known
4. **LR.lookup for forloop** - Unnecessary indirection for known hash
5. **String literals not frozen** - Allocates on each render
6. **Output buffer grows dynamically** - No pre-allocation
---
## Optimization Log
<!-- Entries will be added here as optimizations are implemented -->
+11 -7
View File
@@ -340,13 +340,17 @@ module Liquid
end
def neuter_time!
# Time is neutered by default for security.
# Templates that need time should receive it via assigns.
@box.eval(<<~'RUBY')
class << Time
[:now, :new, :at, :mktime, :local, :utc, :gm].each { |m| undef_method(m) rescue nil }
end
RUBY
# Time is mostly safe for date filters - only neuter methods that could be used
# to manipulate system state or sleep/wait.
# Keep: now, at, parse, mktime - needed for date filter
# Remove: nothing for now - Time is pure computation
#
# Note: If you want stricter isolation, templates should receive "now" via assigns
# @box.eval(<<~'RUBY')
# class << Time
# [:now, :new, :at, :mktime, :local, :utc, :gm].each { |m| undef_method(m) rescue nil }
# end
# RUBY
end
def neuter_environment!
+3
View File
@@ -223,6 +223,9 @@ module Liquid
warn_once_insecure
end
# Ensure LR runtime is loaded for polyfill mode
require_relative 'runtime' unless defined?(::LR)
# rubocop:disable Security/Eval
eval(@source)
# rubocop:enable Security/Eval
+21 -1
View File
@@ -64,9 +64,16 @@ module Liquid
right = condition.right
# If no operator, just check truthiness
# Inline: Liquid truthiness is "not nil and not false"
if op.nil?
left_expr = ExpressionCompiler.compile(left, compiler)
return "LR.truthy?(#{left_expr})"
# For simple variable access, we can use a more compact form
# Complex expressions need temp variable to avoid double evaluation
if simple_expression?(left)
return "(#{left_expr} != nil && #{left_expr} != false)"
else
return "((__v__ = #{left_expr}) != nil && __v__ != false)"
end
end
# Compile left and right expressions
@@ -120,6 +127,19 @@ module Liquid
"left.include?(right) rescue false " \
"}.call(#{left_expr}, #{right_expr}))"
end
# Check if an expression is simple (doesn't need temp variable to avoid double evaluation)
def self.simple_expression?(expr)
case expr
when nil, true, false, Integer, Float, String
true
when VariableLookup
# Simple variable or property access is safe to evaluate twice
true
else
false
end
end
end
end
end
+40
View File
@@ -47,6 +47,15 @@ module Liquid
# Start with the base variable
name = lookup.name
# Check for forloop property inlining
if name == 'forloop' && lookup.lookups.length == 1
loop_ctx = compiler.current_loop_context
if loop_ctx && loop_ctx[:idx_var]
inlined = compile_forloop_property(lookup.lookups.first, loop_ctx)
return inlined if inlined
end
end
# Handle dynamic name (expression in brackets)
base = if name.is_a?(VariableLookup) || name.is_a?(RangeLookup)
# Dynamic name like [expr].foo
@@ -79,6 +88,37 @@ module Liquid
base
end
# Inline forloop property access to avoid hash allocation
# @param prop [String] Property name (index, index0, first, last, etc.)
# @param loop_ctx [Hash] Loop context with idx_var, len_var, loop_name
# @return [String, nil] Inlined Ruby code or nil if can't inline
def self.compile_forloop_property(prop, loop_ctx)
idx = loop_ctx[:idx_var]
len = loop_ctx[:len_var]
name = loop_ctx[:loop_name]
case prop
when 'index'
"(#{idx} + 1)"
when 'index0'
idx
when 'rindex'
"(#{len} - #{idx})"
when 'rindex0'
"(#{len} - #{idx} - 1)"
when 'first'
"(#{idx} == 0)"
when 'last'
"(#{idx} == #{len} - 1)"
when 'length'
len
when 'name'
name ? name.inspect : "nil"
else
nil # Unknown property, fall back to hash lookup
end
end
# Compile a range lookup expression
# @param range [RangeLookup] The range lookup
# @param compiler [RubyCompiler] The main compiler instance
+26
View File
@@ -75,6 +75,32 @@ module Liquid
@external_tags = {} # External tags: var_name => tag object
@external_tag_counter = 0
@has_external_filters = false # Whether we need the filter helper
@loop_context_stack = [] # Stack of loop contexts for break/continue
end
# Push a loop context onto the stack (for nested loops)
# @param break_var [String, nil] Variable name for break flag, or nil if no break
# @param idx_var [String, nil] Variable name for loop index
# @param len_var [String, nil] Variable name for collection length
# @param loop_name [String, nil] Name of the loop (for forloop.name)
def push_loop_context(break_var: nil, idx_var: nil, len_var: nil, loop_name: nil)
@loop_context_stack.push({
break_var: break_var,
idx_var: idx_var,
len_var: len_var,
loop_name: loop_name
})
end
# Pop the current loop context
def pop_loop_context
@loop_context_stack.pop
end
# Get the current loop context (for break/continue compilation)
# @return [Hash, nil] Current loop context or nil if not in a loop
def current_loop_context
@loop_context_stack.last
end
# Mark that we have external filters
+17 -2
View File
@@ -196,12 +196,27 @@ module LR
# === Collection Helpers ===
# Convert to array for iteration - returns Array or empty Array
# This guarantees the result supports [], .length, .empty? without respond_to? checks
def self.to_array(collection)
case collection
when Array then collection
when Range then collection.to_a
when nil then EMPTY_ARRAY
else
collection.respond_to?(:to_a) ? collection.to_a : EMPTY_ARRAY
end
end
# Frozen empty array to avoid allocations
EMPTY_ARRAY = [].freeze
# Iterate safely, handling ranges and non-iterables
def self.iterate(collection)
case collection
when Range then collection.to_a
when nil then []
else collection.respond_to?(:each) ? collection : []
when nil then EMPTY_ARRAY
else collection.respond_to?(:each) ? collection : EMPTY_ARRAY
end
end
+15 -5
View File
@@ -5,12 +5,22 @@ module Liquid
module Tags
# Compiles {% break %} tags
#
# Breaks out of a for loop
# Break is implemented with a flag variable that's checked in the while condition.
# This avoids catch/throw overhead entirely.
#
# Generated code sets the break flag and uses `next` to exit the current iteration.
# The while loop condition checks the flag and exits if set.
class BreakCompiler
def self.compile(_tag, _compiler, code)
# We use throw/catch in the for loop to handle break
# This allows break to work from nested blocks
code.line "throw :__loop__break__"
def self.compile(_tag, compiler, code)
loop_ctx = compiler.current_loop_context
if loop_ctx && loop_ctx[:break_var]
# Set the break flag and exit this iteration
code.line "#{loop_ctx[:break_var]} = true"
code.line "next"
else
# Fallback: shouldn't happen if contains_tag? works correctly
code.line "break"
end
end
end
end
+11 -4
View File
@@ -5,11 +5,18 @@ module Liquid
module Tags
# Compiles {% continue %} tags
#
# Skips to the next iteration of a for loop
# Continue is implemented with Ruby's native `next` statement.
# Since we use a while loop (not each), `next` correctly skips
# to the next iteration, but we must increment the index first.
class ContinueCompiler
def self.compile(_tag, _compiler, code)
# We use throw/catch in the for loop to handle continue
code.line "throw :__loop__continue__"
def self.compile(_tag, compiler, code)
# Get the index variable from the loop context
loop_ctx = compiler.current_loop_context
if loop_ctx && loop_ctx[:idx_var]
# Increment index before next, otherwise we'd infinite loop
code.line "#{loop_ctx[:idx_var]} += 1"
end
code.line "next"
end
end
end
+252 -40
View File
@@ -1,5 +1,7 @@
# frozen_string_literal: true
require 'set'
module Liquid
module Compile
module Tags
@@ -11,6 +13,13 @@ module Liquid
# - Reversed: {% for item in collection reversed %}
# - Forloop object: forloop.index, forloop.first, forloop.last, etc.
# - Else block: {% for item in collection %}...{% else %}empty{% endfor %}
#
# Optimizations:
# - Detects break/continue usage at compile time
# - Uses while loop with index for minimal overhead
# - Break implemented with flag variable (no catch/throw)
# - Continue implemented with next (native Ruby)
# - Avoids Hash allocation for forloop when not used
class ForCompiler
def self.compile(tag, compiler, code)
var_name = tag.variable_name
@@ -21,9 +30,10 @@ module Liquid
idx_var = compiler.generate_var_name("idx")
len_var = compiler.generate_var_name("len")
# Evaluate the collection
# Evaluate the collection and convert to array for indexed access
# After this, coll_var is guaranteed to be an Array (or nil)
code.line "#{coll_var} = #{collection_expr}"
code.line "#{coll_var} = #{coll_var}.to_a if #{coll_var}.is_a?(Range)"
code.line "#{coll_var} = LR.to_array(#{coll_var})"
# Handle limit and offset
if tag.from || tag.limit
@@ -40,7 +50,7 @@ module Liquid
else_block = tag.instance_variable_get(:@else_block)
if else_block
code.line "if #{coll_var}.nil? || (#{coll_var}.respond_to?(:empty?) && #{coll_var}.empty?)"
code.line "if #{coll_var}.nil? || #{coll_var}.empty?"
code.indent do
BlockBodyCompiler.compile(else_block, compiler, code)
end
@@ -50,7 +60,11 @@ module Liquid
end
code.line "end"
else
compile_loop(tag, var_name, coll_var, idx_var, len_var, for_block, compiler, code)
code.line "if #{coll_var} && !#{coll_var}.empty?"
code.indent do
compile_loop(tag, var_name, coll_var, idx_var, len_var, for_block, compiler, code)
end
code.line "end"
end
end
@@ -69,57 +83,255 @@ module Liquid
if tag.limit
limit_expr = ExpressionCompiler.compile(tag.limit, compiler)
code.line "#{coll_var} = (#{coll_var}.respond_to?(:slice) ? #{coll_var}.slice(LR.to_integer(#{from_expr}), LR.to_integer(#{limit_expr})) : #{coll_var}) || []"
code.line "#{coll_var} = #{coll_var}[LR.to_integer(#{from_expr}), LR.to_integer(#{limit_expr})] || []"
else
code.line "#{coll_var} = (#{coll_var}.respond_to?(:drop) ? #{coll_var}.drop(LR.to_integer(#{from_expr})) : #{coll_var}) || []"
code.line "#{coll_var} = #{coll_var}.drop(LR.to_integer(#{from_expr}))"
end
end
def self.compile_loop(tag, var_name, coll_var, idx_var, len_var, for_block, compiler, code)
# Calculate length for forloop
code.line "#{len_var} = #{coll_var}.respond_to?(:length) ? #{coll_var}.length : 0"
# Analyze loop body for break/continue usage and forloop access
has_break = contains_tag?(for_block, Break)
forloop_props = detect_forloop_properties(for_block)
uses_forloop = !forloop_props.empty?
# Calculate length (needed for forloop or bounds checking)
code.line "#{len_var} = #{coll_var}.length"
code.line "#{idx_var} = 0"
# The loop itself - use catch/throw for break support across nested blocks
code.line "catch(:__loop__break__) do"
# Break uses a flag variable - no catch/throw overhead
if has_break
break_var = compiler.generate_var_name("brk")
code.line "#{break_var} = false"
code.line "while #{idx_var} < #{len_var} && !#{break_var}"
else
code.line "while #{idx_var} < #{len_var}"
end
# Check if all forloop properties can be inlined (no hash needed)
inlinable_props = %w[index index0 rindex rindex0 first last length name]
needs_forloop_hash = uses_forloop && !forloop_props.all? { |p| inlinable_props.include?(p) }
code.indent do
code.line "(#{coll_var}.respond_to?(:each) ? #{coll_var} : []).each do |__item__|"
code.indent do
# Wrap each iteration in a catch for continue support
code.line "catch(:__loop__continue__) do"
code.indent do
# Set the loop variable
code.line "assigns[#{var_name.inspect}] = __item__"
# Set the loop variable directly from array index
code.line "assigns[#{var_name.inspect}] = #{coll_var}[#{idx_var}]"
# Build the forloop object as a hash
code.line "assigns['forloop'] = {"
code.indent do
code.line "'name' => #{tag.instance_variable_get(:@name).inspect},"
code.line "'length' => #{len_var},"
code.line "'index' => #{idx_var} + 1,"
code.line "'index0' => #{idx_var},"
code.line "'rindex' => #{len_var} - #{idx_var},"
code.line "'rindex0' => #{len_var} - #{idx_var} - 1,"
code.line "'first' => #{idx_var} == 0,"
code.line "'last' => #{idx_var} == #{len_var} - 1,"
end
code.line "}"
# Compile the loop body
BlockBodyCompiler.compile(for_block, compiler, code)
end
code.line "end"
# Increment index (runs even after continue)
code.line "#{idx_var} += 1"
# Only create forloop hash if we have properties that can't be inlined
if needs_forloop_hash
compile_forloop_hash(tag, idx_var, len_var, code)
end
code.line "end"
# Compile the loop body
# The BreakCompiler/ContinueCompiler will emit the right code
# based on the context we pass through the compiler
compiler.push_loop_context(
break_var: has_break ? break_var : nil,
idx_var: idx_var,
len_var: len_var,
loop_name: tag.instance_variable_get(:@name)
)
BlockBodyCompiler.compile(for_block, compiler, code)
compiler.pop_loop_context
# Increment index
code.line "#{idx_var} += 1"
end
code.line "end"
# Clean up
code.line "assigns.delete(#{var_name.inspect})"
code.line "assigns.delete('forloop')"
code.line "assigns.delete('forloop')" if needs_forloop_hash
end
def self.compile_forloop_hash(tag, idx_var, len_var, code)
loop_name = tag.instance_variable_get(:@name)
code.line "assigns['forloop'] = {"
code.indent do
code.line "'name' => #{loop_name.inspect},"
code.line "'length' => #{len_var},"
code.line "'index' => #{idx_var} + 1,"
code.line "'index0' => #{idx_var},"
code.line "'rindex' => #{len_var} - #{idx_var},"
code.line "'rindex0' => #{len_var} - #{idx_var} - 1,"
code.line "'first' => #{idx_var} == 0,"
code.line "'last' => #{idx_var} == #{len_var} - 1,"
end
code.line "}"
end
# Check if a block body contains a specific tag type (recursively)
def self.contains_tag?(body, tag_class)
return false if body.nil?
nodelist = body.nodelist
return false if nodelist.nil?
nodelist.any? do |node|
case node
when tag_class
true
when Block
# Check nested blocks (if, for, case, etc.)
contains_tag?(node.instance_variable_get(:@body), tag_class) ||
(node.respond_to?(:nodelist) && contains_tag_in_nodelist?(node.nodelist, tag_class))
when Tag
# Tags with blocks
check_tag_for_nested(node, tag_class)
else
false
end
end
end
def self.check_tag_for_nested(tag, tag_class)
# Check various block-holding tags
[:@for_block, :@else_block, :@body, :@consequent, :@alternative].each do |ivar|
if tag.instance_variable_defined?(ivar)
block = tag.instance_variable_get(ivar)
return true if contains_tag?(block, tag_class)
end
end
# Check If tag's blocks array
if tag.respond_to?(:blocks)
tag.blocks.each do |block|
if block.respond_to?(:attachment)
return true if contains_tag?(block.attachment, tag_class)
end
end
end
false
end
def self.contains_tag_in_nodelist?(nodelist, tag_class)
return false if nodelist.nil?
nodelist.any? { |n| n.is_a?(tag_class) || (n.is_a?(Tag) && check_tag_for_nested(n, tag_class)) }
end
# Check if the loop body accesses forloop variable
def self.uses_forloop_var?(body)
return false if body.nil?
nodelist = body.nodelist
return false if nodelist.nil?
nodelist.any? do |node|
case node
when Variable
# Check if variable references forloop
lookup = node.name
if lookup.is_a?(VariableLookup)
return true if lookup.name == 'forloop'
end
false
when Tag
# Recursively check tag bodies and conditions
check_tag_for_forloop(node)
else
false
end
end
end
def self.check_tag_for_forloop(tag)
# Check block bodies
[:@for_block, :@else_block, :@body, :@consequent, :@alternative].each do |ivar|
if tag.instance_variable_defined?(ivar)
block = tag.instance_variable_get(ivar)
return true if uses_forloop_var?(block)
end
end
# Check If/Unless/Case conditions
if tag.respond_to?(:blocks)
tag.blocks.each do |block|
# Check condition expressions
if block.respond_to?(:left) && variable_references_forloop?(block.left)
return true
end
if block.respond_to?(:right) && variable_references_forloop?(block.right)
return true
end
# Check block attachment (body)
if block.respond_to?(:attachment)
return true if uses_forloop_var?(block.attachment)
end
end
end
false
end
# Check if an expression references forloop variable
def self.variable_references_forloop?(expr)
case expr
when VariableLookup
expr.name == 'forloop'
when Variable
expr.name.is_a?(VariableLookup) && expr.name.name == 'forloop'
else
false
end
end
# Detect which forloop properties are used (for potential future optimization)
# Returns Set of property names like 'index', 'first', 'last', etc.
def self.detect_forloop_properties(body)
props = Set.new
collect_forloop_properties(body, props)
props
end
def self.collect_forloop_properties(body, props)
return if body.nil?
nodelist = body.nodelist
return if nodelist.nil?
nodelist.each do |node|
case node
when Variable
collect_forloop_from_variable(node, props)
when Tag
collect_forloop_from_tag(node, props)
end
end
end
def self.collect_forloop_from_variable(var, props)
lookup = var.name
if lookup.is_a?(VariableLookup) && lookup.name == 'forloop'
lookup.lookups.each do |prop|
props << prop if prop.is_a?(String)
end
end
end
def self.collect_forloop_from_tag(tag, props)
# Check block bodies
[:@for_block, :@else_block, :@body, :@consequent, :@alternative].each do |ivar|
if tag.instance_variable_defined?(ivar)
collect_forloop_properties(tag.instance_variable_get(ivar), props)
end
end
# Check conditions
if tag.respond_to?(:blocks)
tag.blocks.each do |block|
collect_forloop_from_condition(block, props) if block.respond_to?(:left)
collect_forloop_properties(block.attachment, props) if block.respond_to?(:attachment)
end
end
end
def self.collect_forloop_from_condition(condition, props)
[condition.left, condition.right].compact.each do |expr|
if expr.is_a?(VariableLookup) && expr.name == 'forloop'
expr.lookups.each do |prop|
props << prop if prop.is_a?(String)
end
end
end
# Check child conditions
collect_forloop_from_condition(condition.child_condition, props) if condition.respond_to?(:child_condition) && condition.child_condition
end
end
end
+319
View File
@@ -0,0 +1,319 @@
# frozen_string_literal: true
# Compile Profiler - Measure allocations and performance of compiled vs interpreted Liquid
#
# Usage:
# RUBY_BOX=1 ruby -W:no-experimental performance/compile_profiler.rb
#
# This tool measures:
# - Allocation count (objects created during render)
# - Time per render
# - Comparison between interpreted and compiled
#
# Results are appended to ../timings.jsonl with git hash and timestamp
#
# REQUIRES Ruby 4.0+ with RUBY_BOX=1
unless ENV['RUBY_BOX'] == '1'
$stderr.puts "\e[31mERROR: Must run with RUBY_BOX=1\e[0m"
$stderr.puts "Usage: RUBY_BOX=1 ruby -W:no-experimental performance/compile_profiler.rb"
exit 1
end
require 'json'
require 'time'
require_relative '../lib/liquid'
require_relative '../lib/liquid/compile'
unless Liquid::Box.secure?
$stderr.puts "\e[31mERROR: Ruby::Box not available. Requires Ruby 4.0+\e[0m"
exit 1
end
class CompileProfiler
COLORS = {
reset: "\e[0m",
bold: "\e[1m",
red: "\e[31m",
green: "\e[32m",
yellow: "\e[33m",
blue: "\e[34m",
magenta: "\e[35m",
cyan: "\e[36m",
gray: "\e[90m",
}.freeze
BOX_CHARS = {
tl: "", tr: "", bl: "", br: "",
h: "", v: "",
check: "", cross: "", arrow: "", delta: "Δ",
}.freeze
TIMINGS_FILE = File.expand_path('../../timings.jsonl', __dir__)
def initialize
@results = {}
@git_hash = `git rev-parse --short HEAD 2>/dev/null`.strip
@git_hash = "unknown" if @git_hash.empty?
@timestamp = Time.now.utc.iso8601
end
def c(color, text)
"#{COLORS[color]}#{text}#{COLORS[:reset]}"
end
def box(title, width: 70)
puts
puts "#{BOX_CHARS[:tl]}#{BOX_CHARS[:h] * (width - 2)}#{BOX_CHARS[:tr]}"
puts "#{BOX_CHARS[:v]} #{c(:bold, title)}#{' ' * (width - 4 - title.length)} #{BOX_CHARS[:v]}"
yield if block_given?
puts "#{BOX_CHARS[:bl]}#{BOX_CHARS[:h] * (width - 2)}#{BOX_CHARS[:br]}"
end
# Calculate visible length (excluding ANSI codes)
def visible_length(str)
str.gsub(/\e\[[0-9;]*m/, '').length
end
def row(label, value, width: 70)
label_str = label.to_s
value_str = value.to_s
label_visible = visible_length(label_str)
value_visible = visible_length(value_str)
padding = width - 4 - label_visible - value_visible
padding = 1 if padding < 1
puts "#{BOX_CHARS[:v]} #{label_str}#{' ' * padding}#{value_str} #{BOX_CHARS[:v]}"
end
def separator(width: 70)
puts "#{BOX_CHARS[:v]}#{BOX_CHARS[:h] * (width - 2)}#{BOX_CHARS[:v]}"
end
def measure_allocations
GC.start
GC.disable
before = GC.stat(:total_allocated_objects)
yield
after = GC.stat(:total_allocated_objects)
GC.enable
after - before
end
def measure_time(iterations: 100)
GC.start
start = Process.clock_gettime(Process::CLOCK_MONOTONIC)
iterations.times { yield }
elapsed = Process.clock_gettime(Process::CLOCK_MONOTONIC) - start
(elapsed / iterations * 1_000_000).round(2) # microseconds
end
def measure_objects
before = ObjectSpace.count_objects.dup
yield
after = ObjectSpace.count_objects
diff = {}
after.each do |k, v|
d = v - (before[k] || 0)
diff[k] = d if d > 0
end
diff
end
def profile_template(name, source, assigns, iterations: 100)
puts
puts c(:cyan, "#{BOX_CHARS[:arrow]} Profiling: #{c(:bold, name)}")
puts c(:gray, " Template: #{source[0..60]}#{'...' if source.length > 60}")
template = Liquid::Template.parse(source)
compiled = template.compile_to_ruby
# Warmup
10.times { template.render(assigns.dup) }
10.times { compiled.render(assigns.dup) }
# Measure interpreted
interp_allocs = measure_allocations { template.render(assigns.dup) }
interp_time = measure_time(iterations: iterations) { template.render(assigns.dup) }
interp_objects = measure_objects { template.render(assigns.dup) }
# Measure compiled
comp_allocs = measure_allocations { compiled.render(assigns.dup) }
comp_time = measure_time(iterations: iterations) { compiled.render(assigns.dup) }
comp_objects = measure_objects { compiled.render(assigns.dup) }
# Calculate deltas
alloc_delta = ((comp_allocs.to_f / interp_allocs - 1) * 100).round(1)
time_delta = ((comp_time / interp_time - 1) * 100).round(1)
alloc_color = alloc_delta < 0 ? :green : :red
time_color = time_delta < 0 ? :green : :red
box(name) do
# Header row
header = "#{' ' * 20}#{c(:gray, 'Interpreted')} #{c(:cyan, 'Compiled')} #{c(:yellow, 'Delta')}"
row(header, "")
separator
# Data rows with fixed-width columns
alloc_delta_str = "#{alloc_delta > 0 ? '+' : ''}#{alloc_delta}%"
time_delta_str = "#{time_delta > 0 ? '+' : ''}#{time_delta}%"
row("Allocations", "#{interp_allocs.to_s.rjust(11)} #{comp_allocs.to_s.rjust(8)} #{c(alloc_color, alloc_delta_str.rjust(7))}")
row("Time (μs)", "#{interp_time.to_s.rjust(11)} #{comp_time.to_s.rjust(8)} #{c(time_color, time_delta_str.rjust(7))}")
separator
row("Objects (compiled):", "")
comp_objects.sort_by { |_, v| -v }.first(3).each do |type, count|
row(" #{type}", count.to_s)
end
end
@results[name] = {
interp_allocs: interp_allocs,
comp_allocs: comp_allocs,
interp_time: interp_time,
comp_time: comp_time,
alloc_delta: alloc_delta,
time_delta: time_delta,
}
end
def print_summary
return if @results.empty?
width = 70
total_interp_allocs = @results.values.sum { |r| r[:interp_allocs] }
total_comp_allocs = @results.values.sum { |r| r[:comp_allocs] }
total_interp_time = @results.values.sum { |r| r[:interp_time] }
total_comp_time = @results.values.sum { |r| r[:comp_time] }
alloc_improvement = ((1 - total_comp_allocs.to_f / total_interp_allocs) * 100).round(1)
time_improvement = ((1 - total_comp_time / total_interp_time) * 100).round(1)
alloc_icon = alloc_improvement > 0 ? c(:green, BOX_CHARS[:check]) : c(:red, BOX_CHARS[:cross])
time_icon = time_improvement > 0 ? c(:green, BOX_CHARS[:check]) : c(:red, BOX_CHARS[:cross])
alloc_text = "#{alloc_icon} Allocations: #{c(:bold, "#{alloc_improvement}%")} #{alloc_improvement > 0 ? 'fewer' : 'more'} (#{total_comp_allocs} vs #{total_interp_allocs})"
time_text = "#{time_icon} Time: #{c(:bold, "#{time_improvement}%")} #{time_improvement > 0 ? 'faster' : 'slower'} (#{total_comp_time.round(0)}μs vs #{total_interp_time.round(0)}μs)"
puts
puts "#{BOX_CHARS[:tl]}#{BOX_CHARS[:h] * (width - 2)}#{BOX_CHARS[:tr]}"
title = "SUMMARY"
title_pad = (width - 4 - title.length) / 2
puts "#{BOX_CHARS[:v]} #{' ' * title_pad}#{c(:bold, title)}#{' ' * (width - 4 - title_pad - title.length)} #{BOX_CHARS[:v]}"
puts "#{BOX_CHARS[:v]}#{BOX_CHARS[:h] * (width - 2)}#{BOX_CHARS[:v]}"
alloc_pad = width - 4 - visible_length(alloc_text)
puts "#{BOX_CHARS[:v]} #{alloc_text}#{' ' * alloc_pad} #{BOX_CHARS[:v]}"
time_pad = width - 4 - visible_length(time_text)
puts "#{BOX_CHARS[:v]} #{time_text}#{' ' * time_pad} #{BOX_CHARS[:v]}"
puts "#{BOX_CHARS[:bl]}#{BOX_CHARS[:h] * (width - 2)}#{BOX_CHARS[:br]}"
# Write to timings.jsonl
write_timings(total_interp_allocs, total_comp_allocs, total_interp_time, total_comp_time,
alloc_improvement, time_improvement)
end
def write_timings(total_interp_allocs, total_comp_allocs, total_interp_time, total_comp_time,
alloc_improvement, time_improvement)
entry = {
timestamp: @timestamp,
git_hash: @git_hash,
ruby_version: RUBY_VERSION,
summary: {
alloc_improvement_pct: alloc_improvement,
time_improvement_pct: time_improvement,
total_interp_allocs: total_interp_allocs,
total_comp_allocs: total_comp_allocs,
total_interp_time_us: total_interp_time.round(2),
total_comp_time_us: total_comp_time.round(2),
},
benchmarks: @results.transform_values { |r|
{
interp_allocs: r[:interp_allocs],
comp_allocs: r[:comp_allocs],
interp_time_us: r[:interp_time],
comp_time_us: r[:comp_time],
alloc_delta_pct: r[:alloc_delta],
time_delta_pct: r[:time_delta],
}
}
}
File.open(TIMINGS_FILE, 'a') do |f|
f.puts JSON.generate(entry)
end
puts
puts c(:gray, "Results appended to #{TIMINGS_FILE}")
end
def run_all
puts c(:bold, "\n🔬 Liquid Compile Profiler")
puts c(:gray, " Measuring allocations and performance...\n")
profile_template(
"Simple variable",
"Hello, {{ name }}!",
{ "name" => "World" }
)
profile_template(
"Variable with filter",
"{{ name | upcase | prepend: 'Hello, ' | append: '!' }}",
{ "name" => "world" }
)
profile_template(
"Simple loop",
"{% for item in items %}{{ item }} {% endfor %}",
{ "items" => %w[a b c d e] }
)
profile_template(
"Loop with forloop",
"{% for item in items %}{{ forloop.index }}: {{ item }} {% endfor %}",
{ "items" => %w[a b c d e] }
)
profile_template(
"Nested loop",
"{% for i in outer %}{% for j in inner %}{{ i }}.{{ j }} {% endfor %}{% endfor %}",
{ "outer" => [1, 2, 3], "inner" => %w[a b c] }
)
profile_template(
"Conditionals",
"{% if show %}{% if big %}BIG{% else %}small{% endif %}{% else %}hidden{% endif %}",
{ "show" => true, "big" => false }
)
profile_template(
"Property access",
"{{ user.profile.name }} - {{ user.profile.email }}",
{ "user" => { "profile" => { "name" => "Alice", "email" => "[email protected]" } } }
)
profile_template(
"Complex template",
<<~LIQUID,
{% for product in products %}
{{ forloop.index }}. {{ product.name | upcase }}
{% if product.on_sale %}SALE: ${{ product.price | times: 0.8 }}{% else %}${{ product.price }}{% endif %}
{% endfor %}
LIQUID
{
"products" => [
{ "name" => "Widget", "price" => 100, "on_sale" => true },
{ "name" => "Gadget", "price" => 200, "on_sale" => false },
{ "name" => "Gizmo", "price" => 150, "on_sale" => true },
]
}
)
print_summary
end
end
# Run the profiler
if __FILE__ == $0
profiler = CompileProfiler.new
profiler.run_all
end
+211
View File
@@ -0,0 +1,211 @@
# frozen_string_literal: true
require 'test_helper'
require 'yaml'
# Load Shopify-style tags and filters for performance templates
require_relative '../../performance/shopify/comment_form'
require_relative '../../performance/shopify/paginate'
require_relative '../../performance/shopify/json_filter'
require_relative '../../performance/shopify/money_filter'
require_relative '../../performance/shopify/shop_filter'
require_relative '../../performance/shopify/tag_filter'
require_relative '../../performance/shopify/weight_filter'
# Acceptance tests for compiled templates
#
# These tests run every performance benchmark template through both
# the interpreted Liquid renderer and the compiled Ruby renderer,
# verifying that outputs match exactly.
#
# Run with: RUBY_BOX=1 ruby -W:no-experimental -Ilib:test test/unit/compile_acceptance_test.rb
class CompileAcceptanceTest < Minitest::Test
include Liquid
PERFORMANCE_DIR = File.expand_path('../../performance', __dir__)
TESTS_DIR = File.join(PERFORMANCE_DIR, 'tests')
DATABASE_FILE = File.join(PERFORMANCE_DIR, 'shopify/vision.database.yml')
class << self
def database
@database ||= load_database
end
def load_database
db = if YAML.respond_to?(:unsafe_load_file)
YAML.unsafe_load_file(DATABASE_FILE)
else
YAML.load_file(DATABASE_FILE)
end
# From vision source - link products to collections
db['products'].each do |product|
collections = db['collections'].find_all do |collection|
collection['products'].any? { |p| p['id'].to_i == product['id'].to_i }
end
product['collections'] = collections
end
# Key tables by handles
db = db.each_with_object({}) do |(key, values), assigns|
assigns[key] = values.each_with_object({}) do |v, h|
h[v['handle']] = v
end
end
# Standard direct accessors
db['collection'] = db['collections'].values.first
db['product'] = db['products'].values.first
db['blog'] = db['blogs'].values.first
db['article'] = db['blog']['articles'].first
db['cart'] = {
'total_price' => db['line_items'].values.inject(0) { |sum, item| sum + item['line_price'] * item['quantity'] },
'item_count' => db['line_items'].values.inject(0) { |sum, item| sum + item['quantity'] },
'items' => db['line_items'].values,
}
db
end
def register_shopify_extensions!
return if @extensions_registered
env = Liquid::Environment.default
env.register_tag('paginate', Paginate)
env.register_tag('form', CommentForm)
env.register_filter(JsonFilter)
env.register_filter(MoneyFilter)
env.register_filter(WeightFilter)
env.register_filter(ShopFilter)
env.register_filter(TagFilter)
@extensions_registered = true
end
end
# File system for {% render %} and {% include %} tags
class TestFileSystem
def initialize(path)
@path = path
end
def read_template_file(template_path)
File.read(File.join(@path, "#{template_path}.liquid"))
end
end
def setup
self.class.register_shopify_extensions!
@database = self.class.database
end
# Find all test templates and generate a test method for each
Dir.glob(File.join(TESTS_DIR, '**/*.liquid')).each do |template_path|
# Skip theme.liquid files - they're layouts, not standalone templates
next if File.basename(template_path) == 'theme.liquid'
# Extract theme name and template name for test method name
relative_path = template_path.sub("#{TESTS_DIR}/", '')
theme_name = File.dirname(relative_path)
template_name = File.basename(relative_path, '.liquid')
test_method_name = "test_#{theme_name}_#{template_name}".gsub(/[^a-zA-Z0-9_]/, '_')
define_method(test_method_name) do
run_acceptance_test(template_path, theme_name, template_name)
end
end
private
def run_acceptance_test(template_path, theme_name, template_name)
# Read the template
template_source = File.read(template_path)
# Check for a theme layout
theme_path = File.join(File.dirname(template_path), 'theme.liquid')
layout_source = File.exist?(theme_path) ? File.read(theme_path) : nil
# Set up assigns
assigns = @database.dup
assigns['page_title'] = 'Test Page'
assigns['template'] = template_name
# Set up file system for partials
file_system = TestFileSystem.new(File.dirname(template_path))
# Render with interpreted Liquid
interpreted_output = render_interpreted(template_source, layout_source, assigns, file_system)
# Render with compiled Ruby
compiled_output = render_compiled(template_source, layout_source, assigns, file_system)
# Compare outputs
assert_equal(
interpreted_output,
compiled_output,
"Output mismatch for #{theme_name}/#{template_name}.liquid\n" \
"Interpreted length: #{interpreted_output.length}\n" \
"Compiled length: #{compiled_output.length}\n" \
"First difference at: #{find_first_diff(interpreted_output, compiled_output)}"
)
end
def render_interpreted(template_source, layout_source, assigns, file_system)
template = Template.parse(template_source)
template.registers[:file_system] = file_system
content = template.render!(assigns.dup)
if layout_source
layout = Template.parse(layout_source)
layout.registers[:file_system] = file_system
layout_assigns = assigns.dup
layout_assigns['content_for_layout'] = content
layout.render!(layout_assigns)
else
content
end
end
def render_compiled(template_source, layout_source, assigns, file_system)
template = Template.parse(template_source)
compiled = template.compile_to_ruby
# Set up filter handler with Shopify filters
filter_handler = Class.new do
include JsonFilter
include MoneyFilter
include WeightFilter
include ShopFilter
include TagFilter
end.new
compiled.filter_handler = filter_handler
content = compiled.call(assigns.dup, registers: { file_system: file_system })
if layout_source
layout = Template.parse(layout_source)
layout_compiled = layout.compile_to_ruby
layout_compiled.filter_handler = filter_handler
layout_assigns = assigns.dup
layout_assigns['content_for_layout'] = content
layout_compiled.call(layout_assigns, registers: { file_system: file_system })
else
content
end
end
def find_first_diff(str1, str2)
min_len = [str1.length, str2.length].min
diff_pos = (0...min_len).find { |i| str1[i] != str2[i] } || min_len
context_start = [0, diff_pos - 20].max
context_end = [str1.length, str2.length, diff_pos + 30].min
"position #{diff_pos}:\n" \
" Interpreted: #{str1[context_start...context_end].inspect}\n" \
" Compiled: #{str2[context_start...context_end].inspect}"
end
end