unicode width updates

- 0xFE0F (VS16) needs to move the cursor and clear the overlapping cell
  when it's combined with a previous cell, which adds some complexity if
  the next print was going to need to wrap
- tmux uses a hardcoded list of codepoints that are always forced wide
  since they are typically rendered as wide even if the unicode standard
  or wcwidth() typically considers them narrow. add that table here as
  well so tmux and mosh can stay in sync for these codepoints
- unassigned codepoints should default to a width of 1 instead of not
  being printed at all, which allows unknown codepoints to still
  possibly be rendered

Author: jdrouhard

src/terminal/terminal.cc

@@ -134,8 +134,26 @@ void Emulator::print( const Parser::Print* act )
       }
       if ( !combining_cell->full() ) {
         combining_cell->append( ch );
-        if ( ch == 0xFE0F ) {
+        // VS16 causes the previous codepoint to  be rendered as its emoji representation
+        // which could cause it to change from 1 to 2 characters wide
+        if ( ch == 0xFE0F && !combining_cell->get_wide() ) {
+          // have to move this emoji to the next line
+          if ( fb.ds.auto_wrap_mode && fb.ds.next_print_will_wrap ) {
+            fb.get_mutable_row( -1 )->set_wrap( false );
+            fb.ds.move_col( 0 );
+            fb.move_rows_autoscroll( 1 );
+            *fb.get_mutable_cell() = *combining_cell;
+            fb.reset_cell( combining_cell );
+            fb.ds.move_col( 1, true, true );
+            combining_cell = fb.get_combining_cell();
+          }
           combining_cell->set_wide( true );
+          if ( fb.ds.insert_mode ) {
+            fb.insert_cell( fb.ds.get_cursor_row(), fb.ds.get_cursor_col() );
+          } else if ( fb.ds.get_cursor_col() < fb.ds.get_width() ) {
+            fb.reset_cell( fb.get_mutable_cell() );
+          }
+          fb.ds.move_col( 1, true, true );
         }
       }
     } break;

src/util/char_utils.h

@@ -2,33 +2,60 @@
 #define CHAR_UTILS_HPP
 
 #include "widechar_width.h"
+#include <algorithm>
+#include <array>
 
 static int mosh_wcwidth( uint32_t c )
 {
-  int width = widechar_wcwidth( c );
-  if ( width >= 0 ) {
-    return width;
-  }
+  int result = [c]() {
+    int width = widechar_wcwidth( c );
+    if ( width >= 0 ) {
+      return width;
+    }
+
+    /* https://github.com/ridiculousfish/widecharwidth/tree/master#c-usage */
+    switch ( width ) {
+      case widechar_nonprint:
+        return -1;
+      case widechar_combining:
+        return 0;
+      case widechar_ambiguous:
+        return 1;
+      case widechar_private_use:
+        return 1;
+      case widechar_unassigned:
+        return 1;
+      case widechar_non_character:
+        return -1;
+      case widechar_widened_in_9:
+        return 2;
+      default:
+        return -1;
+    }
+  }();
+
+  static constexpr auto force_wide = std::array {
+    0x0261D, 0x026F9, 0x0270A, 0x0270B, 0x0270C, 0x0270D, 0x1F1E6, 0x1F1E7, 0x1F1E8, 0x1F1E9, 0x1F1EA, 0x1F1EB,
+    0x1F1EC, 0x1F1ED, 0x1F1EE, 0x1F1EF, 0x1F1F0, 0x1F1F1, 0x1F1F2, 0x1F1F3, 0x1F1F4, 0x1F1F5, 0x1F1F6, 0x1F1F7,
+    0x1F1F8, 0x1F1F9, 0x1F1FA, 0x1F1FB, 0x1F1FC, 0x1F1FD, 0x1F1FE, 0x1F1FF, 0x1F385, 0x1F3C2, 0x1F3C3, 0x1F3C4,
+    0x1F3C7, 0x1F3CA, 0x1F3CB, 0x1F3CC, 0x1F3FB, 0x1F3FC, 0x1F3FD, 0x1F3FE, 0x1F3FF, 0x1F442, 0x1F443, 0x1F446,
+    0x1F447, 0x1F448, 0x1F449, 0x1F44A, 0x1F44B, 0x1F44C, 0x1F44D, 0x1F44E, 0x1F44F, 0x1F450, 0x1F466, 0x1F467,
+    0x1F468, 0x1F469, 0x1F46B, 0x1F46C, 0x1F46D, 0x1F46E, 0x1F470, 0x1F471, 0x1F472, 0x1F473, 0x1F474, 0x1F475,
+    0x1F476, 0x1F477, 0x1F478, 0x1F47C, 0x1F481, 0x1F482, 0x1F483, 0x1F485, 0x1F486, 0x1F487, 0x1F48F, 0x1F491,
+    0x1F4AA, 0x1F574, 0x1F575, 0x1F57A, 0x1F590, 0x1F595, 0x1F596, 0x1F645, 0x1F646, 0x1F647, 0x1F64B, 0x1F64C,
+    0x1F64D, 0x1F64E, 0x1F64F, 0x1F6A3, 0x1F6B4, 0x1F6B5, 0x1F6B6, 0x1F6C0, 0x1F6CC, 0x1F90C, 0x1F90F, 0x1F918,
+    0x1F919, 0x1F91A, 0x1F91B, 0x1F91C, 0x1F91D, 0x1F91E, 0x1F91F, 0x1F926, 0x1F930, 0x1F931, 0x1F932, 0x1F933,
+    0x1F934, 0x1F935, 0x1F936, 0x1F937, 0x1F938, 0x1F939, 0x1F93D, 0x1F93E, 0x1F977, 0x1F9B5, 0x1F9B6, 0x1F9B8,
+    0x1F9B9, 0x1F9BB, 0x1F9CD, 0x1F9CE, 0x1F9CF, 0x1F9D1, 0x1F9D2, 0x1F9D3, 0x1F9D4, 0x1F9D5, 0x1F9D6, 0x1F9D7,
+    0x1F9D8, 0x1F9D9, 0x1F9DA, 0x1F9DB, 0x1F9DC, 0x1F9DD, 0x1FAC3, 0x1FAC4, 0x1FAC5, 0x1FAF0, 0x1FAF1, 0x1FAF2,
+    0x1FAF3, 0x1FAF4, 0x1FAF5, 0x1FAF6, 0x1FAF7, 0x1FAF8 };
 
-  /* https://github.com/ridiculousfish/widecharwidth/tree/master#c-usage */
-  switch ( width ) {
-    case widechar_nonprint:
-      return -1;
-    case widechar_combining:
-      return 0;
-    case widechar_ambiguous:
-      return 1;
-    case widechar_private_use:
-      return 1;
-    case widechar_unassigned:
-      return -1;
-    case widechar_non_character:
-      return -1;
-    case widechar_widened_in_9:
+  if ( result == 1 ) {
+    if ( std::binary_search( force_wide.begin(), force_wide.end(), c ) ) {
       return 2;
-    default:
-      return -1;
+    }
   }
+  return result;
 }
 
 #endif