commit 7824318: [Minor] Add util.if_utf_mixed_script to lua

[Miecio Za]

Author: Miecio Za
Date: 2019-03-18 19:56:57 +0100
URL: https://github.com/rspamd/rspamd/commit/7824318dca65d223df397329b3eabde86c3fde02

[Minor] Add util.if_utf_mixed_script to lua
Add new function which implements PoC for chekcing
mixed script in utf string. Behaviour is similar
to single string spoof detection in libicu before
version 58

---
 src/lua/lua_util.c | 51 +++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 51 insertions(+)

diff --git a/src/lua/lua_util.c b/src/lua/lua_util.c
index 1a37eaef6..881257ed3 100644
--- a/src/lua/lua_util.c
+++ b/src/lua/lua_util.c
@@ -29,6 +29,7 @@
 #include <zlib.h>
 
 #include "unicode/uspoof.h"
+#include "unicode/uscript.h"
 
 /***
  * @module rspamd_util
@@ -393,6 +394,14 @@ LUA_FUNCTION_DEF (util, normalize_prob);
  */
 LUA_FUNCTION_DEF (util, is_utf_spoofed);
 
+/**
+* @function util.is_utf_mixed_script(str)
+* Returns true if a string contains mixed unicode scripts
+* @param {string} String to check
+* @return {boolean} true if a string contains chars with mixed unicode script
+*/
+LUA_FUNCTION_DEF (util, is_utf_mixed_script);
+
 /**
 * @function util.is_utf_outside_range(str, range_start, range_end)
 * Returns true if a string contains chars outside range
@@ -633,6 +642,7 @@ static const struct luaL_reg utillib_f[] = {
 	LUA_INTERFACE_DEF (util, caseless_hash),
 	LUA_INTERFACE_DEF (util, caseless_hash_fast),
 	LUA_INTERFACE_DEF (util, is_utf_spoofed),
+	LUA_INTERFACE_DEF (util, is_utf_mixed_script),
 	LUA_INTERFACE_DEF (util, is_utf_outside_range),
 	LUA_INTERFACE_DEF (util, get_string_stats),
 	LUA_INTERFACE_DEF (util, is_valid_utf8),
@@ -2498,6 +2508,47 @@ lua_util_is_utf_spoofed (lua_State *L)
 	return nres;
 }
 
+static gint
+lua_util_is_utf_mixed_script(lua_State *L)
+{
+	LUA_TRACE_POINT;
+	gsize len_of_string;
+	const gchar *end, *string_to_check = lua_tolstring (L, 1, &len_of_string);
+	UScriptCode last_script_code = USCRIPT_INVALID_CODE;
+	UErrorCode uc_err = U_ZERO_ERROR;
+
+	if (string_to_check && g_utf8_validate (string_to_check, len_of_string, &end)) {
+		len_of_string = g_utf8_strlen (string_to_check, len_of_string);
+
+		for(; *string_to_check; string_to_check = g_utf8_next_char(string_to_check)){
+			gunichar char_to_check = g_utf8_get_char(string_to_check);
+			UScriptCode current_script_code = uscript_getScript(char_to_check, &uc_err);
+			if (uc_err != U_ZERO_ERROR){
+				msg_err ("cannot get unicode script for character, error: %s", u_errorName (uc_err));
+				lua_pushboolean (L, false);
+				return 1;
+			}
+			if ( current_script_code != USCRIPT_COMMON && current_script_code != USCRIPT_INHERITED ){
+				if (last_script_code == USCRIPT_INVALID_CODE ){
+					last_script_code = current_script_code;
+				} else {
+					if ( last_script_code != current_script_code ){
+						lua_pushboolean (L, true);
+						return 1;
+					}
+				}
+			}
+		}
+	}
+	else {
+		return luaL_error (L, "invalid arguments");
+	}
+
+	lua_pushboolean (L, false);
+
+	return 1;
+}
+
 static gint
 lua_util_get_string_stats (lua_State *L)
 {