diff --git a/asdl/front_end_test.py b/asdl/front_end_test.py
index 382ea6b61d..3502462e58 100755
--- a/asdl/front_end_test.py
+++ b/asdl/front_end_test.py
@@ -41,6 +41,14 @@ def testSharedVariantCode(self):
     self.assertEqual(1001, expr_e.DoubleQuoted)
     self.assertEqual(expr_e.DoubleQuoted, word_part_e.DoubleQuoted)
 
+    d = double_quoted(5, ['foo', 'bar'])
+    d.PrettyPrint()
+    print()
+
+    b = expr.Binary(d, d)
+    b.PrettyPrint()
+
+
 
 if __name__ == '__main__':
   unittest.main()
diff --git a/asdl/gen_python.py b/asdl/gen_python.py
index a1d7fe2bad..7ea7d4c55a 100755
--- a/asdl/gen_python.py
+++ b/asdl/gen_python.py
@@ -319,10 +319,9 @@ def VisitCompoundSum(self, sum, sum_name, depth):
 
   def VisitProduct(self, product, name, depth):
     self._shared_type_tags[name] = self._product_counter
-    self._product_counter += 1
-
     self._GenClass(product, product.attributes, name, 'runtime.CompoundObj',
-                   depth)
+                   depth, tag_num=self._product_counter)
+    self._product_counter += 1
 
   def EmitFooter(self):
     pass
diff --git a/frontend/syntax.asdl b/frontend/syntax.asdl
index b5c4624b92..365a13df97 100644
--- a/frontend/syntax.asdl
+++ b/frontend/syntax.asdl
@@ -97,6 +97,8 @@ module syntax
   | Slice(arith_expr? begin, arith_expr? length)
   attributes (int* spids)
 
+  double_quoted = (token left, word_part* parts) attributes (int* spids)
+
   word_part = 
     ArrayLiteral(word* words)
     -- alternating key and value (saving some space)
@@ -105,7 +107,7 @@ module syntax
     -- escaped case is separate so the evaluator doesn't have to check token ID
   | EscapedLiteral(token token)
   | SingleQuoted(token left, token* tokens)
-  | DoubleQuoted(token left, word_part* parts)
+  | DoubleQuoted %double_quoted
   | SimpleVarSub(token token)
 
     -- TODO: Add Oil expressions
@@ -337,7 +339,7 @@ module syntax
   | SingleQuoted(token left, token* tokens)
     -- NOTE: Duplicates word_part.DoubleQuoted
     -- Not all word_part are valid here!
-  | DoubleQuoted(token left, word_part* parts)
+  | DoubleQuoted %double_quoted
 
   | Unary(token op, expr child)
   | Binary(token op, expr left, expr right)
diff --git a/frontend/syntax_abbrev.py b/frontend/syntax_abbrev.py
index 5915ab9c40..64ff417571 100644
--- a/frontend/syntax_abbrev.py
+++ b/frontend/syntax_abbrev.py
@@ -28,6 +28,20 @@ def _token(obj):
   return p_node
 
 
+def _double_quoted(obj):
+  # type: (double_quoted) -> PrettyNode
+  if obj.left.id != Id.Left_DoubleQuote:
+    return None  # Fall back on obj._AbbreviatedTree()
+
+  p_node = runtime.PrettyNode()
+  p_node.abbrev = True
+  p_node.node_type = 'DQ'
+
+  for part in obj.parts:
+    p_node.unnamed_fields.append(part.AbbreviatedTree())
+  return p_node
+
+
 def _word_part__Literal(obj):
   # type: (word_part__Literal) -> PrettyNode
   p_node = runtime.PrettyNode()
@@ -59,20 +73,6 @@ def _word_part__BracedVarSub(obj):
   return p_node
 
 
-def _word_part__DoubleQuoted(obj):
-  # type: (word_part__DoubleQuoted) -> PrettyNode
-  if obj.left.id != Id.Left_DoubleQuote:
-    return None  # Fall back on obj._AbbreviatedTree()
-
-  p_node = runtime.PrettyNode()
-  p_node.abbrev = True
-  p_node.node_type = 'DQ'
-
-  for part in obj.parts:
-    p_node.unnamed_fields.append(part.AbbreviatedTree())
-  return p_node
-
-
 def _word_part__SingleQuoted(obj):
   # type: (word_part__SingleQuoted) -> PrettyNode
 
diff --git a/oil_lang/expr_parse.py b/oil_lang/expr_parse.py
index fa25aad2ad..3a2a5c9c5b 100644
--- a/oil_lang/expr_parse.py
+++ b/oil_lang/expr_parse.py
@@ -6,7 +6,8 @@
 import sys
 
 from _devbuild.gen.syntax_asdl import (
-    token, word__Token, word__Compound, word_part, word_part_t, expr
+    token, double_quoted, word__Token, word__Compound, word_part, word_part_t,
+    expr
 )
 from _devbuild.gen.id_kind_asdl import Id, Kind
 from _devbuild.gen.types_asdl import lex_mode_e
@@ -312,7 +313,7 @@ def _PushOilTokens(parse_ctx, gr, p, lex):
 
       parts = []  # type: List[word_part_t]
       last_token = w_parser.ReadDoubleQuoted(left_token, parts)
-      expr_dq_part = expr.DoubleQuoted(left_token, parts)
+      expr_dq_part = double_quoted(left_token, parts)
 
       typ = Id.Expr_CastedDummy.enum_id
       opaque = cast(token, expr_dq_part)  # HACK for expr_to_ast
diff --git a/oil_lang/expr_to_ast.py b/oil_lang/expr_to_ast.py
index 78841912aa..0f24d7fb1f 100644
--- a/oil_lang/expr_to_ast.py
+++ b/oil_lang/expr_to_ast.py
@@ -5,9 +5,9 @@
 
 from _devbuild.gen.id_kind_asdl import Id
 from _devbuild.gen.syntax_asdl import (
-    token, command, command__VarDecl,
-    expr, expr_t,
-    expr__DoubleQuoted, expr__SingleQuoted, expr__Dict, expr__BracedVarSub,
+    token, double_quoted,
+    command, command__VarDecl,
+    expr, expr_t, expr__SingleQuoted, expr__Dict, expr__BracedVarSub,
     expr_context_e, regex, regex_t,
     word_t, word_part__CommandSub, 
     param, type_expr_t,
@@ -496,7 +496,7 @@ def Expr(self, pnode):
         return part
 
       elif typ == grammar_nt.dq_string:
-        dq_part = cast(expr__DoubleQuoted, children[1].tok)
+        dq_part = cast(double_quoted, children[1].tok)
         return dq_part
 
       elif typ == grammar_nt.sq_string:
diff --git a/osh/word_.py b/osh/word_.py
index 14ccc53dbb..7c13667d5b 100644
--- a/osh/word_.py
+++ b/osh/word_.py
@@ -5,10 +5,11 @@
 from _devbuild.gen.id_kind_asdl import (Id, Kind, Id_t, Kind_t)
 from _devbuild.gen.syntax_asdl import (
     token,
+    double_quoted,
     word_part, word_part_t, word_part_e,
     word_part__ArrayLiteral, word_part__AssocArrayLiteral,
     word_part__Literal, word_part__EscapedLiteral,
-    word_part__SingleQuoted, word_part__DoubleQuoted,
+    word_part__SingleQuoted,
     word_part__SimpleVarSub, word_part__BracedVarSub, word_part__TildeSub,
     word_part__CommandSub, word_part__ArithSub, word_part__BracedTuple,
     word_part__ExtGlob, word_part__Splice, word_part__FuncCall,
@@ -84,7 +85,7 @@ def _EvalWordPart(part):
     s = ''.join(t.val for t in part.tokens)
     return True, s, True
 
-  elif isinstance(part, word_part__DoubleQuoted):
+  elif isinstance(part, double_quoted):
     ret = ''
     for p in part.parts:
       ok, s, _ = _EvalWordPart(p)
@@ -145,7 +146,7 @@ def LeftMostSpanForPart(part):
   elif isinstance(part, word_part__SingleQuoted):
     return part.left.span_id  # single quote location
 
-  elif isinstance(part, word_part__DoubleQuoted):
+  elif isinstance(part, double_quoted):
     return part.left.span_id  # double quote location
 
   elif isinstance(part, word_part__SimpleVarSub):
@@ -201,7 +202,7 @@ def _RightMostSpanForPart(part):
   elif isinstance(part, word_part__SingleQuoted):
     return part.spids[1]  # right '
 
-  elif isinstance(part, word_part__DoubleQuoted):
+  elif isinstance(part, double_quoted):
     return part.spids[1]  # right "
 
   elif isinstance(part, word_part__SimpleVarSub):
diff --git a/osh/word_parse.py b/osh/word_parse.py
index 6bab2a97a1..f1a2b54e67 100644
--- a/osh/word_parse.py
+++ b/osh/word_parse.py
@@ -51,14 +51,14 @@
 from _devbuild.gen.id_kind_asdl import Id, Kind, Id_t
 from _devbuild.gen.types_asdl import lex_mode_t, lex_mode_e
 from _devbuild.gen.syntax_asdl import (
-    token, arith_expr_t,
+    token, double_quoted, arith_expr_t,
     suffix_op, suffix_op_t, suffix_op__Slice, suffix_op__PatSub,
     bracket_op, bracket_op_t,
 
     word, word_t, word__Compound, word__Token,
     word_part, word_part_t,
     word_part__Literal, word_part__BracedVarSub, word_part__SingleQuoted,
-    word_part__ArithSub, word_part__DoubleQuoted, word_part__CommandSub,
+    word_part__ArithSub, word_part__CommandSub,
     word_part__ExtGlob,
 
     command, command_t, command__ForExpr,
@@ -687,7 +687,7 @@ def _ReadLikeDQ(self, left_dq_token, out_parts):
     # Return nothing, since we appended to 'out_parts'
 
   def _ReadDoubleQuoted(self):
-    # type: () -> word_part__DoubleQuoted
+    # type: () -> double_quoted
     """
     Args:
       eof_type: for stopping at }, Id.Lit_RBrace
@@ -699,7 +699,7 @@ def _ReadDoubleQuoted(self):
     parts = []  # type: List[word_part_t]
     self._ReadLikeDQ(left_dq_token, parts)
 
-    dq_part = word_part.DoubleQuoted(left_dq_token, parts)
+    dq_part = double_quoted(left_dq_token, parts)
     dq_part.spids.append(left_dq_token.span_id)  # Left ", sort of redundant
     dq_part.spids.append(self.cur_token.span_id)  # Right "
     return dq_part