fix(str): remove support for unneeded word boundaries

CHANGES.md

@@ -14,6 +14,8 @@ Unreleased
 * Fix [Re.Pcre.split]. Regression introduced in 1.12 and a previous bug with
   [Re.Pcre.split] (#538).
 
+* Avoid parsing unnecessary patterns supported `Re.Emacs` in `Re.Str` (#563)
+
 1.13.1 (30-Sep-2024)
 --------------------
 

lib/emacs.ml

@@ -31,7 +31,7 @@ let by_code f c c' =
   Char.chr (f c c')
 ;;
 
-let parse s =
+let parse ~emacs_only s =
   let buf = Parse_buffer.create s in
   let accept = Parse_buffer.accept buf in
   let eos () = Parse_buffer.eos buf in
@@ -72,19 +72,19 @@ let parse s =
         let r = regexp () in
         if not (Parse_buffer.accept_s buf {|\)|}) then raise Parse_error;
         Re.group r)
-      else if accept '`'
+      else if emacs_only && accept '`'
       then Re.bos
-      else if accept '\''
+      else if emacs_only && accept '\''
       then Re.eos
       else if accept '='
       then Re.start
       else if accept 'b'
       then Re.alt [ Re.bow; Re.eow ]
-      else if accept 'B'
+      else if emacs_only && accept 'B'
       then Re.not_boundary
-      else if accept '<'
+      else if emacs_only && accept '<'
       then Re.bow
-      else if accept '>'
+      else if emacs_only && accept '>'
       then Re.eow
       else if accept 'w'
       then Re.alt [ Re.alnum; Re.char '_' ]
@@ -95,7 +95,7 @@ let parse s =
         match get () with
         | ('*' | '+' | '?' | '[' | ']' | '.' | '^' | '$' | '\\') as c -> Re.char c
         | '0' .. '9' -> raise Not_supported
-        | _ -> raise Parse_error)
+        | c -> if emacs_only then raise Parse_error else Re.char c)
     else (
       if eos () then raise Parse_error;
       match get () with
@@ -125,7 +125,12 @@ let parse s =
 ;;
 
 let re ?(case = true) s =
-  let r = parse s in
+  let r = parse s ~emacs_only:true in
+  if case then r else Re.no_case r
+;;
+
+let re_no_emacs ~case s =
+  let r = parse s ~emacs_only:false in
   if case then r else Re.no_case r
 ;;
 

lib/emacs.mli

@@ -37,3 +37,5 @@ val compile : Core.t -> Core.re
 
 (** Same as [Core.compile] *)
 val compile_pat : ?case:bool -> string -> Core.re
+
+val re_no_emacs : case:bool -> string -> Core.t

lib/str.ml

@@ -30,7 +30,7 @@ type regexp =
   }
 
 let compile_regexp s c =
-  let re = Emacs.re ~case:(not c) s in
+  let re = Emacs.re_no_emacs ~case:(not c) s in
   { mtch = lazy (Compile.compile (Ast.seq [ Ast.start; re ]))
   ; srch = lazy (Compile.compile re)
   }

lib_test/str/test_str.ml

@@ -27,6 +27,15 @@ module Test_matches (R : Str_intf) = struct
     with
     | Not_found -> None
   ;;
+
+  let eq_match' ?(pos = 0) ?(case = true) r s =
+    let pat = if case then R.regexp r else R.regexp_case_fold r in
+    try
+      ignore (R.string_match pat s pos);
+      Some (groups ())
+    with
+    | Not_found -> None
+  ;;
 end
 
 module T_str = Test_matches (Str)
@@ -42,6 +51,16 @@ let eq_match ?pos ?case r s =
     ()
 ;;
 
+let eq_match' ?pos ?case r s =
+  expect_equal_app
+    ~msg:(str_printer s)
+    ~printer:(opt_printer (list_printer ofs_printer))
+    (fun () -> T_str.eq_match' ?pos ?case r s)
+    ()
+    (fun () -> T_re.eq_match' ?pos ?case r s)
+    ()
+;;
+
 let split_result_conv =
   List.map (function
     | Str.Delim x -> Re.Str.Delim x
@@ -206,6 +225,14 @@ let _ =
     eq_match "[^0-9a-z]+" "A:Z+";
     eq_match "[^0-9a-z]+" "0";
     eq_match "[^0-9a-z]+" "a");
+  (* Word modifiers *)
+  expect_pass "word boundaries" (fun () ->
+    eq_match' "\\bfoo" "foo";
+    eq_match' "\\<foo" "foo";
+    eq_match' "foo\\>" "foo";
+    eq_match' "z\\Bfoo" "zfoo";
+    eq_match' "\\`foo" "foo";
+    eq_match' "foo\\'" "foo");
   (* Case modifiers *)
   expect_pass "no_case" (fun () ->
     eq_match ~case:false "abc" "abc";