Faster Expression parser / Tokenizer with StringScanner

Gemfile

@@ -13,6 +13,7 @@ group :benchmark, :test do
   gem 'benchmark-ips'
   gem 'memory_profiler'
   gem 'terminal-table'
+  gem "lru_redux"
 
   install_if -> { RUBY_PLATFORM !~ /mingw|mswin|java/ && RUBY_ENGINE != 'truffleruby' } do
     gem 'stackprof'

Rakefile

@@ -71,7 +71,7 @@ end
 
 namespace :benchmark do
   desc "Run the liquid benchmark with lax parsing"
-  task :run do
+  task :lax do
     ruby "./performance/benchmark.rb lax"
   end
 
@@ -80,11 +80,30 @@ namespace :benchmark do
     ruby "./performance/benchmark.rb strict"
   end
 
+  desc "Run the liquid benchmark with both lax and strict parsing"
+  task run: [:lax, :strict]
+
   desc "Run unit benchmarks"
-  task :unit do
-    Dir["./performance/unit/*_benchmark.rb"].each do |file|
-      puts "🧪 Running #{file}"
-      ruby file
+  namespace :unit do
+    task :all do
+      Dir["./performance/unit/*_benchmark.rb"].each do |file|
+        puts "🧪 Running #{file}"
+        ruby file
+      end
+    end
+
+    task :lexer do
+      Dir["./performance/unit/lexer_benchmark.rb"].each do |file|
+        puts "🧪 Running #{file}"
+        ruby file
+      end
+    end
+
+    task :expression do
+      Dir["./performance/unit/expression_benchmark.rb"].each do |file|
+        puts "🧪 Running #{file}"
+        ruby file
+      end
     end
   end
 end

lib/liquid.rb

@@ -21,6 +21,8 @@
 # OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
 # WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
+require "strscan"
+
 module Liquid
   FilterSeparator             = /\|/
   ArgumentSeparator           = ','
@@ -44,6 +46,7 @@ module Liquid
   VariableParser              = /\[(?>[^\[\]]+|\g<0>)*\]|#{VariableSegment}+\??/o
 
   RAISE_EXCEPTION_LAMBDA = ->(_e) { raise }
+  HAS_STRING_SCANNER_SCAN_BYTE = StringScanner.instance_methods.include?(:scan_byte)
 end
 
 require "liquid/version"
@@ -68,7 +71,6 @@ module Liquid
 require 'liquid/errors'
 require 'liquid/interrupts'
 require 'liquid/strainer_template'
-require 'liquid/expression'
 require 'liquid/context'
 require 'liquid/tag'
 require 'liquid/block_body'
@@ -77,6 +79,7 @@ module Liquid
 require 'liquid/variable_lookup'
 require 'liquid/range_lookup'
 require 'liquid/resource_limits'
+require 'liquid/expression'
 require 'liquid/template'
 require 'liquid/condition'
 require 'liquid/utils'

lib/liquid/context.rb

@@ -1,5 +1,7 @@
 # frozen_string_literal: true
 
+require "lru_redux"
+
 module Liquid
   # Context keeps the variable stack and resolves variables, as well as keywords
   #
@@ -39,6 +41,11 @@ def initialize(environments = {}, outer_scope = {}, registers = {}, rethrow_erro
       @filters             = []
       @global_filter       = nil
       @disabled_tags       = {}
+      @expression_cache    = LruRedux::ThreadSafeCache.new(1000)
+
+      # Instead of constructing new StringScanner objects for each Expression parse,
+      # we recycle the same one.
+      @string_scanner = StringScanner.new("")
 
       @registers.static[:cached_partials] ||= {}
       @registers.static[:file_system] ||= environment.file_system
@@ -176,7 +183,7 @@ def []=(key, value)
     # Example:
     #   products == empty #=> products.empty?
     def [](expression)
-      evaluate(Expression.parse(expression))
+      evaluate(Expression.parse(expression, @string_scanner, @expression_cache))
     end
 
     def key?(key)

lib/liquid/expression.rb

@@ -1,5 +1,7 @@
 # frozen_string_literal: true
 
+require "lru_redux"
+
 module Liquid
   class Expression
     LITERALS = {
@@ -10,37 +12,106 @@ class Expression
       'true' => true,
       'false' => false,
       'blank' => '',
-      'empty' => ''
+      'empty' => '',
+      # in lax mode, minus sign can be a VariableLookup
+      # For simplicity and performace, we treat it like a literal
+      '-' => VariableLookup.parse("-", nil).freeze,
     }.freeze
 
-    INTEGERS_REGEX       = /\A(-?\d+)\z/
-    FLOATS_REGEX         = /\A(-?\d[\d\.]+)\z/
+    DOT = ".".ord
+    ZERO = "0".ord
+    NINE = "9".ord
+    DASH = "-".ord
 
     # Use an atomic group (?>...) to avoid pathological backtracing from
     # malicious input as described in https://github.com/Shopify/liquid/issues/1357
-    RANGES_REGEX         = /\A\(\s*(?>(\S+)\s*\.\.)\s*(\S+)\s*\)\z/
+    RANGES_REGEX = /\A\(\s*(?>(\S+)\s*\.\.)\s*(\S+)\s*\)\z/
+    INTEGER_REGEX = /\A(-?\d+)\z/
+    FLOAT_REGEX = /\A(-?\d+)\.\d+\z/
+
+    class << self
+      def parse(markup, ss = StringScanner.new(""), cache = nil)
+        return unless markup
+
+        markup = markup.strip # markup can be a frozen string
 
-    def self.parse(markup)
-      return nil unless markup
+        if (markup.start_with?('"') && markup.end_with?('"')) ||
+          (markup.start_with?("'") && markup.end_with?("'"))
+          return markup[1..-2]
+        elsif LITERALS.key?(markup)
+          return LITERALS[markup]
+        end
+
+        # Cache only exists during parsing
+        if cache
+          return cache[markup] if cache.key?(markup)
 
-      markup = markup.strip
-      if (markup.start_with?('"') && markup.end_with?('"')) ||
-         (markup.start_with?("'") && markup.end_with?("'"))
-        return markup[1..-2]
+          cache[markup] = inner_parse(markup, ss, cache).freeze
+        else
+          inner_parse(markup, ss, nil).freeze
+        end
       end
 
-      case markup
-      when INTEGERS_REGEX
-        Regexp.last_match(1).to_i
-      when RANGES_REGEX
-        RangeLookup.parse(Regexp.last_match(1), Regexp.last_match(2))
-      when FLOATS_REGEX
-        Regexp.last_match(1).to_f
-      else
-        if LITERALS.key?(markup)
-          LITERALS[markup]
+      def inner_parse(markup, ss, cache)
+        if (markup.start_with?("(") && markup.end_with?(")")) && markup =~ RANGES_REGEX
+          return RangeLookup.parse(
+            Regexp.last_match(1),
+            Regexp.last_match(2),
+            ss,
+            cache,
+          )
+        end
+
+        if (num = parse_number(markup, ss))
+          num
+        else
+          VariableLookup.parse(markup, ss, cache)
+        end
+      end
+
+      def parse_number(markup, ss)
+        # check if the markup is simple integer or float
+        case markup
+        when INTEGER_REGEX
+          return Integer(markup, 10)
+        when FLOAT_REGEX
+          return markup.to_f
+        end
+
+        ss.string = markup
+        # the first byte must be a digit, a period, or  a dash
+        byte = ss.scan_byte
+
+        return false if byte != DASH && byte != DOT && (byte < ZERO || byte > NINE)
+
+        # The markup could be a float with multiple dots
+        first_dot_pos = nil
+        num_end_pos = nil
+
+        while (byte = ss.scan_byte)
+          return false if byte != DOT && (byte < ZERO || byte > NINE)
+
+          # we found our number and now we are just scanning the rest of the string
+          next if num_end_pos
+
+          if byte == DOT
+            if first_dot_pos.nil?
+              first_dot_pos = ss.pos
+            else
+              # we found another dot, so we know that the number ends here
+              num_end_pos = ss.pos - 1
+            end
+          end
+        end
+
+        num_end_pos = markup.length if ss.eos?
+
+        if num_end_pos
+          # number ends with a number "123.123"
+          markup.byteslice(0, num_end_pos).to_f
         else
-          VariableLookup.parse(markup)
+          # number ends with a dot "123."
+          markup.byteslice(0, first_dot_pos).to_f
         end
       end
     end