diff --git a/spec/tartrazine_spec.cr b/spec/tartrazine_spec.cr
index 42309a7..7b564c2 100644
--- a/spec/tartrazine_spec.cr
+++ b/spec/tartrazine_spec.cr
@@ -74,7 +74,7 @@ end
 
 # Helper that creates lexer and tokenizes
 def tokenize(lexer_name, text)
-  lexer = Tartrazine::Lexer.from_xml(File.read("lexers/#{lexer_name}.xml"))
+  lexer = Tartrazine.get_lexer(lexer_name)
   collapse_tokens(lexer.tokenize(text))
 end
 
diff --git a/src/actions.cr b/src/actions.cr
index a77f4f9..72eb9f7 100644
--- a/src/actions.cr
+++ b/src/actions.cr
@@ -66,8 +66,8 @@ module Tartrazine
         #
         raise Exception.new "Can't have a token without a match" if match.nil?
 
-        # Each group matches an action
-
+        # Each group matches an action. If the group match is empty,
+        # the action is skipped.
         result = [] of Token
         @actions.each_with_index do |e, i|
           next if match[i + 1]?.nil?
@@ -79,7 +79,7 @@ module Tartrazine
         return [] of Token if match.nil?
         lexer_name = xml["lexer"].downcase
         # pp! "to tokenize:", match[match_group]
-        LEXERS[lexer_name].tokenize(match[match_group], usingself: true)
+        Tartrazine.get_lexer(lexer_name).tokenize(match[match_group], usingself: true)
       when "usingself"
         # Shunt to another copy of this lexer
         return [] of Token if match.nil?
diff --git a/src/tartrazine.cr b/src/tartrazine.cr
index 49a6b41..6434b86 100644
--- a/src/tartrazine.cr
+++ b/src/tartrazine.cr
@@ -13,6 +13,10 @@ module Tartrazine
   # For explanations on what actions, transformers, etc do
   # the Pygments documentation is a good place to start.
   # https://pygments.org/docs/lexerdevelopment/
+
+  # A Lexer state. A state has a name and a list of rules.
+  # The state machine has a state stack containing references
+  # to states to decide which rules to apply.
   class State
     property name : String = ""
     property rules = [] of Rule
@@ -25,10 +29,9 @@ module Tartrazine
     end
   end
 
+  # A token, the output of the tokenizer
   alias Token = NamedTuple(type: String, value: String)
 
-  LEXERS = {} of String => Tartrazine::Lexer
-
   class Lexer
     property config = {
       name:             "",
@@ -135,22 +138,9 @@ module Tartrazine
       l
     end
   end
-end
 
-# Try loading all lexers
-
-lexers = Tartrazine::LEXERS
-
-Dir.glob("lexers/*.xml").each do |fname|
-  begin
-    l = Tartrazine::Lexer.from_xml(File.read(fname))
-  rescue ex : Exception
-    # p! ex
-    next
-  end
-  lexers[l.config[:name].downcase] = l
-  l.config[:aliases].each do |key|
-    lexers[key.downcase] = l
+  def self.get_lexer(name : String) : Lexer
+    Lexer.from_xml(File.read("lexers/#{name}.xml"))
   end
 end
 
@@ -166,29 +156,3 @@ end
 macro xml_to_a(node, name)
 {{node}}.children.select{|n| n.name == "{{name}}".lstrip("_")}.map {|n| n.content.to_s}
 end
-
-# # #<Regex::Error:Regex match error: match limit exceeded>
-# next if testname == "tests/fortran/test_string_cataback.txt"
-
-# # Difference is different unicode representation of a string literal
-# next if testname == "tests/java/test_string_literals.txt"
-# next if testname == "tests/systemd/example1.txt"
-# next if testname == "tests/json/test_strings.txt"
-
-# # Tartrazine agrees with pygments, disagrees with chroma
-# next if testname == "tests/java/test_default.txt"
-# next if testname == "tests/java/test_numeric_literals.txt"
-# next if testname == "tests/java/test_multiline_string.txt"
-
-# # Tartrazine disagrees with pygments and chroma, but it's fine
-# next if testname == "tests/php/test_string_escaping_run.txt"
-
-# # Chroma's output is bad, but so is Tartrazine's
-# next if "tests/html/javascript_unclosed.txt" == testname
-
-# # KNOWN BAD -- TO FIX
-# next if "tests/html/css_backtracking.txt" == testname
-# next if "tests/php/anonymous_class.txt" == testname
-# next if "tests/c/test_string_resembling_decl_end.txt" == testname
-# next if "tests/mcfunction/data.txt" == testname
-# next if "tests/mcfunction/selectors.txt" == testname