feat: raw identifier syntax

mlx/lexer.mll

@@ -355,6 +355,7 @@ let kwdopchar =
 
 let ident = (lowercase | uppercase) identchar*
 let extattrident = ident ('.' ident)*
+let raw_ident_escape = "\\#"
 
 let decimal_literal =
   ['0'-'9'] ['0'-'9' '_']*
@@ -396,6 +397,8 @@ rule token = parse
   | ".~"
       { error lexbuf
           (Reserved_sequence (".~", Some "is reserved for use in MetaOCaml")) }
+  | "~" raw_ident_escape (lowercase identchar * as name) ':'
+      { LABEL name }
   | "~" (lowercase identchar * as name) ':'
       { check_label_name lexbuf name;
         LABEL name }
@@ -404,17 +407,23 @@ rule token = parse
         LABEL name }
   | "?"
       { QUESTION }
+  | "?" raw_ident_escape (lowercase identchar * as name) ':'
+      { OPTLABEL name }
   | "?" (lowercase identchar * as name) ':'
       { check_label_name lexbuf name;
         OPTLABEL name }
   | "?" (lowercase_latin1 identchar_latin1 * as name) ':'
       { warn_latin1 lexbuf;
         OPTLABEL name }
+  | raw_ident_escape (lowercase identchar * as name)
+      { LIDENT name }
   | lowercase identchar * as name
       { try Hashtbl.find keyword_table name
         with Not_found -> LIDENT name }
   | lowercase_latin1 identchar_latin1 * as name
       { warn_latin1 lexbuf; LIDENT name }
+  | "<" raw_ident_escape (lowercase identchar * as name)
+      { JSX_LIDENT name }
   | "<" (lowercase identchar * as name)
       { JSX_LIDENT name }
   | "<" "/" (lowercase identchar * as name)

ocamlmerlin_mlx/ocaml/preprocess/lexer_raw.mll

@@ -369,6 +369,7 @@ let kwdopchar =
 
 let ident = (lowercase | uppercase) identchar*
 let extattrident = ident ('.' ident)*
+let raw_ident_escape = "\\#"
 
 let decimal_literal =
   ['0'-'9'] ['0'-'9' '_']*
@@ -426,17 +427,23 @@ rule token state = parse
       { fail lexbuf
           (Reserved_sequence (".~", Some "is reserved for use in MetaOCaml")) }
       *)
+  | "~" raw_ident_escape (lowercase identchar * as name) ':'
+      { return (LABEL name) }
   | "~" (lowercase identchar * as name) ':'
       { lABEL (check_label_name lexbuf name) }
   | "~" (lowercase_latin1 identchar_latin1 * as name) ':'
       { warn_latin1 lexbuf;
         return (LABEL name) }
   | "?"
       { return QUESTION }
+  | "?" raw_ident_escape (lowercase identchar * as name) ':'
+      { return (OPTLABEL name) }
   | "?" (lowercase identchar * as name) ':'
       { oPTLABEL (check_label_name lexbuf name) }
   | "?" (lowercase_latin1 identchar_latin1 * as name) ':'
       { warn_latin1 lexbuf; return (OPTLABEL name) }
+  | raw_ident_escape (lowercase identchar * as name)
+      { return (LIDENT name) }
   | lowercase identchar * as name
     { return (try Hashtbl.find state.keywords name
               with Not_found ->
@@ -445,6 +452,8 @@ rule token state = parse
                 LIDENT name) }
   | lowercase_latin1 identchar_latin1 * as name
       { warn_latin1 lexbuf; return (LIDENT name) }
+  | "<" raw_ident_escape (lowercase identchar * as name)
+      { return (JSX_LIDENT name) }
   | "<" (lowercase identchar * as name)
       { return (JSX_LIDENT name) }
   | "<" "/" (lowercase identchar * as name)

test/mlx.t

@@ -140,3 +140,73 @@ We have a lexer hack to parse [<element and [<Element as JSX:
   let _ = [ (M.element () ~children:[ 1 ] [@JSX]) ]
   MERLIN
   let _ = [ (M.element () ~children:[ 1 ] [@JSX]) ]
+
+Raw identifiers (\#foo) bypass the keyword table.
+Non-keyword raw identifiers round-trip through the full pipeline:
+
+  $ echo 'let _ = \#foo' | ./mlx
+  BATCH
+  let _ = foo
+  MERLIN
+  let _ = foo
+  $ echo 'let _ = \#bar' | ./mlx
+  BATCH
+  let _ = bar
+  MERLIN
+  let _ = bar
+
+Keyword raw identifiers are verified via tokenization (the OCaml printer
+does not emit \# escapes, so the text round-trip breaks for keywords):
+
+  $ echo 'let x = `\#lazy' | mlx-pp -print-tokens
+  let
+  x
+  =
+  `
+  \#lazy
+  $ echo 'let x = `\#true' | mlx-pp -print-tokens
+  let
+  x
+  =
+  `
+  \#true
+  $ echo 'let x = `\#false' | mlx-pp -print-tokens
+  let
+  x
+  =
+  `
+  \#false
+  $ echo 'let f ~\#true:x = x' | mlx-pp -print-tokens
+  let
+  f
+  ~\#true:
+  x
+  =
+  x
+  $ echo 'let f ?\#true:(x = true) = x' | mlx-pp -print-tokens
+  let
+  f
+  ?\#true:
+  (
+  x
+  =
+  true
+  )
+  =
+  x
+  $ echo 'let _ = <\#lazy />' | mlx-pp -print-tokens
+  let
+  _
+  =
+  <\#lazy
+  />
+  $ echo 'let _ = <element loading=`\#lazy />' | mlx-pp -print-tokens
+  let
+  _
+  =
+  <element
+  loading
+  =
+  `
+  \#lazy
+  />