Make normal filtering of plain ASCII lines faster

This patch adds a field lines_not_ascii to the MenuState structure. The
nth entry is 0 unless the nth member of MenuState.lines has a non-ascii
codepoint in it.

All comparison functions (menu_match_cb type) take an additional
argument to tell them if the thing they are matching is not_ascii. They
can use this to determine whether to collate and case-fold the
input (for non-ascii strings), or whether to use strstr/strcasestr (for
ascii strings).

The change is not currently implemented for flex matching, due to my
laziness. However, it should be a simple enough matter to add.

For my large input of 400,000 lines, this reduces typical filtering time
to about ten microseconds from about 2 seconds.

include/helper.h

@@ -102,7 +102,7 @@ int find_arg ( const char * const key );
  *
  * @returns 1 when matches, 0 otherwise
  */
-int token_match ( char **tokens, const char *input, int case_sensitive,
+int token_match ( char **tokens, const char *input, int not_ascii, int case_sensitive,
                   __attribute__( ( unused ) ) unsigned int index,
                   __attribute__( ( unused ) ) Switcher * data );
 
@@ -152,4 +152,11 @@ char helper_parse_char ( const char *arg );
  * Set the application arguments.
  */
 void cmd_set_arguments ( int argc, char **argv );
+
+/**
+ * @param str a UTF8 string
+ * @return 1 if the string contains any non-ascii codepoints
+ */
+int is_not_ascii ( const char *str );
+
 #endif // ROFI_HELPER_H

include/rofi.h

@@ -78,7 +78,7 @@ typedef enum
  *
  * @returns 1 when it matches, 0 if not.
  */
-typedef int ( *menu_match_cb )( char **tokens, const char *input, int case_sensitive, unsigned int index, Switcher *data );
+typedef int ( *menu_match_cb )( char **tokens, const char *input, int not_ascii, int case_sensitive, unsigned int index, Switcher *data );
 
 /**
  * @param sw the Switcher to show.

source/dialogs/combi.c

@@ -169,7 +169,7 @@ static SwitcherMode combi_mode_result ( int mretv, char **input, unsigned int se
     }
     return MODE_EXIT;
 }
-static int combi_mode_match ( char **tokens, const char *input,
+static int combi_mode_match ( char **tokens, const char *input, int not_ascii,
                               int case_sensitive, unsigned int index, Switcher *sw )
 {
     CombiModePrivateData *pd = sw->private_data;
@@ -178,13 +178,13 @@ static int combi_mode_match ( char **tokens, const char *input,
         if ( index >= pd->starts[i] && index < ( pd->starts[i] + pd->lengths[i] ) ) {
             if ( tokens && input[0] && tokens[0][0] == '!' ) {
                 if ( tokens[0][1] == pd->switchers[i]->name[0] ) {
-                    return pd->switchers[i]->token_match ( &tokens[1], input, case_sensitive,
+                    return pd->switchers[i]->token_match ( &tokens[1], input, not_ascii, case_sensitive,
                                                            index - pd->starts[i], pd->switchers[i] );
                 }
                 return 0;
             }
             else {
-                return pd->switchers[i]->token_match ( tokens, input, case_sensitive,
+                return pd->switchers[i]->token_match ( tokens, input, not_ascii, case_sensitive,
                                                        index - pd->starts[i], pd->switchers[i] );
             }
         }

source/dialogs/dmenu.c

@@ -316,7 +316,7 @@ int dmenu_switcher_dialog ( void )
         char         **tokens = tokenize ( select, config.case_sensitive );
         unsigned int i        = 0;
         for ( i = 0; i < cmd_list_length; i++ ) {
-            if ( token_match ( tokens, cmd_list[i], config.case_sensitive, 0, NULL ) ) {
+            if ( token_match ( tokens, cmd_list[i], is_not_ascii(cmd_list[i]), config.case_sensitive, 0, NULL ) ) {
                 pd->selected_line = i;
                 break;
             }

source/dialogs/window.c

@@ -322,6 +322,7 @@ typedef struct _SwitcherModePrivateData
 } SwitcherModePrivateData;
 
 static int window_match ( char **tokens, __attribute__( ( unused ) ) const char *input,
+                          __attribute__( ( unused) ) int not_ascii,
                           int case_sensitive, unsigned int index, Switcher *sw )
 {
     SwitcherModePrivateData *rmpd = (SwitcherModePrivateData *) sw->private_data;
@@ -338,19 +339,19 @@ static int window_match ( char **tokens, __attribute__( ( unused ) ) const char
             // e.g. when searching 'title element' and 'class element'
             char *ftokens[2] = { tokens[j], NULL };
             if ( !test && c->title[0] != '\0' ) {
-                test = token_match ( ftokens, c->title, case_sensitive, 0, NULL );
+                test = token_match ( ftokens, c->title, is_not_ascii(c->title), case_sensitive, 0, NULL );
             }
 
             if ( !test && c->class[0] != '\0' ) {
-                test = token_match ( ftokens, c->class, case_sensitive, 0, NULL );
+                test = token_match ( ftokens, c->class, is_not_ascii(c->title), case_sensitive, 0, NULL );
             }
 
             if ( !test && c->role[0] != '\0' ) {
-                test = token_match ( ftokens, c->role, case_sensitive, 0, NULL );
+                test = token_match ( ftokens, c->role, is_not_ascii(c->title), case_sensitive, 0, NULL );
             }
 
             if ( !test && c->name[0] != '\0' ) {
-                test = token_match ( ftokens, c->name, case_sensitive, 0, NULL );
+                test = token_match ( ftokens, c->name, is_not_ascii(c->title), case_sensitive, 0, NULL );
             }
 
             if ( test == 0 ) {

source/helper.c

@@ -310,11 +310,13 @@ int find_arg_char ( const char * const key, char *val )
  * Shared 'token_match' function.
  * Matches tokenized.
  */
-static int fuzzy_token_match ( char **tokens, const char *input, int case_sensitive )
+static int fuzzy_token_match ( char **tokens, const char *input, __attribute__( (unused) ) int not_ascii,  int case_sensitive )
 {
     int  match  = 1;
     char *compk = token_collate_key ( input, case_sensitive );
     // Do a tokenized match.
+    // TODO: this doesn't work for unicode input, because it may split a codepoint which is over two bytes.
+    // TODO this does not use the non-ascii speed-up either.
     if ( tokens ) {
         for ( int j = 0; match && tokens[j]; j++ ) {
             char *t        = compk;
@@ -331,28 +333,33 @@ static int fuzzy_token_match ( char **tokens, const char *input, int case_sensit
     g_free ( compk );
     return match;
 }
-static int normal_token_match ( char **tokens, const char *input, int case_sensitive )
+static int normal_token_match ( char **tokens, const char *input, int not_ascii, int case_sensitive )
 {
     int  match  = 1;
-    char *compk = token_collate_key ( input, case_sensitive );
+    char *compk = not_ascii ? token_collate_key ( input, case_sensitive ) : (char *) input;
 
     // Do a tokenized match.
+
     if ( tokens ) {
-        for ( int j = 0; match && tokens[j]; j++ ) {
+      char *(*comparison)(const char *, const char *);
-            match = ( strstr ( compk, tokens[j] ) != NULL );
+      comparison = (case_sensitive || not_ascii) ? strstr : strcasestr;
-        }
+      for ( int j = 0; match && tokens[j]; j++ ) {
+        match = (comparison( compk, tokens[j] ) != NULL );
+      }
     }
-    g_free ( compk );
+
+    if (not_ascii) g_free ( compk );
+
     return match;
 }
-int token_match ( char **tokens, const char *input, int case_sensitive,
+int token_match ( char **tokens, const char *input, int not_ascii, int case_sensitive,
                   __attribute__( ( unused ) ) unsigned int index,
                   __attribute__( ( unused ) ) Switcher *data )
 {
     if ( config.fuzzy ) {
-        return fuzzy_token_match ( tokens, input, case_sensitive );
+        return fuzzy_token_match ( tokens, input, not_ascii, case_sensitive );
     }
-    return normal_token_match ( tokens, input, case_sensitive );
+    return normal_token_match ( tokens, input, not_ascii, case_sensitive );
 }
 
 int execute_generator ( const char * cmd )
@@ -478,3 +485,12 @@ void config_sanity_check (  )
         config.menu_bg_alt = config.menu_bg;
     }
 }
+
+int is_not_ascii ( const char * str )
+{
+   while (*str > 0) {
+     str++;
+   }
+   if (*str) return 1;
+   return 0;
+}

source/rofi.c

@@ -244,6 +244,7 @@ typedef struct MenuState
     unsigned int *selected_line;
     MenuReturn   retv;
     char         **lines;
+    int          *lines_not_ascii;
     int          line_height;
 }MenuState;
 
@@ -307,6 +308,8 @@ static void menu_free_state ( MenuState *state )
     g_free ( state->boxes );
     g_free ( state->line_map );
     g_free ( state->distance );
+
+    g_free ( state->lines_not_ascii );
 }
 
 /**
@@ -724,7 +727,7 @@ static void menu_refilter ( MenuState *state )
 
         // input changed
         for ( unsigned int i = 0; i < state->num_lines; i++ ) {
-            int match = state->sw->token_match ( tokens, state->lines[i], config.case_sensitive, i, state->sw );
+            int match = state->sw->token_match ( tokens, state->lines[i], state->lines_not_ascii[i], config.case_sensitive, i, state->sw );
 
             // If each token was matched, add it to list.
             if ( match ) {
@@ -749,6 +752,7 @@ static void menu_refilter ( MenuState *state )
         }
         state->filtered_lines = state->num_lines;
     }
+
     state->selected = MIN ( state->selected, state->filtered_lines - 1 );
 
     if ( config.auto_select == TRUE && state->filtered_lines == 1 && state->num_lines > 1 ) {
@@ -1013,6 +1017,14 @@ MenuReturn menu ( Switcher *sw, char **input, char *prompt, unsigned int *select
     };
     // Request the lines to show.
     state.lines = sw->get_data ( &( state.num_lines ), sw );
+    state.lines_not_ascii = g_malloc0_n( state.num_lines, sizeof( int ) );
+
+    // find out which lines contain non-ascii codepoints, so we can be faster in some cases.
+
+    for (unsigned int line = 0; state.lines[line]; line++) {
+      state.lines_not_ascii[line] = is_not_ascii(state.lines[line]);
+    }
+
     if ( next_pos ) {
         *next_pos = *selected_line;
     }