From 17f635af618463c68e219fee73e74821b6903c7b Mon Sep 17 00:00:00 2001
From: Jarvis Carroll <jarviscarroll@qpq.swiss>
Date: Tue, 3 Feb 2026 00:41:00 +0000
Subject: [PATCH] Parse long hex escape codes

This doesn't work super consistently in the compiler, for codepoints above 127, but it should work fine for us, so, oh well!
---
 src/hz_sophia.erl | 43 +++++++++++++++++++++++++++++++------------
 1 file changed, 31 insertions(+), 12 deletions(-)

diff --git a/src/hz_sophia.erl b/src/hz_sophia.erl
index 5017a92..32bbde4 100644
--- a/src/hz_sophia.erl
+++ b/src/hz_sophia.erl
@@ -103,12 +103,13 @@ reverse_combine_nibbles([D1], Acc) ->
 reverse_combine_nibbles([], Acc) ->
     Acc.
 
-string_token(Start, {tk, Row, Col}, [$\\, $x, A, B | Rest], SourceChars, Value) ->
-    case escape_hex_code(A, B) of
-        {ok, ByteVal} ->
-            string_token(Start, {tk, Row + 4, Col}, Rest, [B, A, $x, $\ | SourceChars], <<Value/binary, ByteVal>>);
-        error ->
-            {error, {invalid_escape_code, [$\\, $x, A, B], Row, Col}}
+string_token(Start, {tk, Row, Col}, "\\x" ++ String, SourceChars, Value) ->
+    case escape_hex_code({tk, Row, Col}, {tk, Row + 2, Col}, String, "x\\" ++ SourceChars) of
+        {ok, {Codepoint, NewSourceChars, NewTk, NewString}} ->
+            NewValue = <<Value/binary, Codepoint/utf8>>,
+            string_token(Start, NewTk, NewString, NewSourceChars, NewValue);
+        {error, Reason} ->
+            {error, Reason}
     end;
 string_token(Start, {tk, Row, Col}, [$\\, C | Rest], SourceChars, Value) ->
     case escape_char(C) of
@@ -122,16 +123,34 @@ string_token({tk, _, Start}, {tk, Row, End}, [$" | Rest], SourceChars, Value) ->
     Token = {string, SourceStr, Value, Row, Start, End},
     {ok, {Token, {tk, Row, End}, Rest}};
 string_token(Start, {tk, Row, Col}, [C | Rest], SourceChars, Value) ->
-    string_token(Start, {tk, Row + 1, Col}, Rest, [C | SourceChars], <<Value/binary, C>>).
+    % TODO: ERTS probably had to convert this FROM utf8 at some point, so why
+    % bother, if we need to convert it back? I guess we could accept iolists if
+    % we really wanted to waste time on this point...
+    string_token(Start, {tk, Row + 1, Col}, Rest, [C | SourceChars], <<Value/binary, C/utf8>>).
 
-escape_hex_code(A, B) when ?IS_HEX(A), ?IS_HEX(B) ->
+escape_hex_code(Start, {tk, Row, Col}, "{" ++ String, SourceChars) ->
+    escape_long_hex_code(Start, {tk, Row + 1, Col}, String, "{" ++ SourceChars, 0);
+escape_hex_code(_, {tk, Row, Col}, [A, B | String], SourceChars) when ?IS_HEX(A), ?IS_HEX(B) ->
     % As of writing this, the Sophia compiler will convert this byte from
     % extended ASCII to unicode... But it really shouldn't. The literal parser
     % does what the compiler should do.
     Byte = convert_digit(A) * 16 + convert_digit(B),
-    {ok, Byte};
-escape_hex_code(_, _) ->
-    error.
+    {ok, {Byte, [B, A | SourceChars], {tk, Row + 2, Col}, String}};
+escape_hex_code({tk, Row1, Col1}, _, _, _) ->
+    {error, {invalid_escape_code, "\\x", Row1, Col1}}.
+
+escape_long_hex_code(_, {tk, Row, Col}, "}" ++ String, SourceChars, Value) ->
+    {ok, {Value, "}" ++ SourceChars, {tk, Row + 1, Col}, String}};
+escape_long_hex_code(Start, {tk, Row, Col}, [C | String], SourceChars, Value) when ?IS_HEX(C) ->
+    NewSourceChars = [C | SourceChars],
+    NewValue = 16 * Value + convert_digit(C),
+    escape_long_hex_code(Start, {tk, Row + 1, Col}, String, NewSourceChars, NewValue);
+escape_long_hex_code(_, {tk, Row, Col}, [C | _], _, _) ->
+    {error, {invalid_hexadecimal, [C], Row, Col}};
+escape_long_hex_code(_, Tk, [], SourceChars, Value) ->
+    % Just return as if the escape code were closed, and let the string parser
+    % produce an unclosed string error instead.
+    {ok, {Value, SourceChars, Tk, []}}.
 
 escape_char($b)  -> {ok, $\b};
 escape_char($e)  -> {ok, $\e};
@@ -747,7 +766,7 @@ anon_types_test() ->
 string_escape_codes_test() ->
     check_parser("\"  \\b\\e\\f\\n\\r\\t\\v\\\"\\\\  \""),
     check_parser("\"\\x00\\x11\\x77\\x4a\\x4A\""),
-    check_parser("\"\\x{7F}\\x{07F}\\x{007F}\\x{0007F}\""),
+    check_parser("\"\\x{0}\\x{7}\\x{7F}\\x{07F}\\x{007F}\\x{0007F}\\x{0000007F}\""),
     ok.
 
 records_test() ->