From 9c2a7522054b15ccd95c0c45e2a566376a037b72 Mon Sep 17 00:00:00 2001
From: Jesse Luehrs <doy@tozt.net>
Date: Tue, 3 May 2016 04:22:09 -0400
Subject: recognize most emoji as wide

---
 Makefile            |  3 +-
 src/screen.c        |  2 +-
 src/unicode-extra.c | 95 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 src/unicode-extra.h |  8 +++++
 src/vt100.h         |  1 +
 5 files changed, 107 insertions(+), 2 deletions(-)
 create mode 100644 src/unicode-extra.c
 create mode 100644 src/unicode-extra.h

diff --git a/Makefile b/Makefile
index de9518b..b1a6484 100644
--- a/Makefile
+++ b/Makefile
@@ -3,7 +3,8 @@ SOUT     = libvt100.a
 BUILD    = build/
 SRC      = src/
 OBJ      = $(BUILD)parser.o \
-	   $(BUILD)screen.o
+	   $(BUILD)screen.o \
+	   $(BUILD)unicode-extra.o
 LIBS     = glib-2.0
 CFLAGS  ?= -g -Wall -Wextra -Werror
 LDFLAGS ?= -g -Wall -Wextra -Werror
diff --git a/src/screen.c b/src/screen.c
index 3faa070..c3c6af7 100644
--- a/src/screen.c
+++ b/src/screen.c
@@ -210,7 +210,7 @@ void vt100_screen_show_string_utf8(VT100Screen *vt, char *buf, size_t len)
 
         uc = g_utf8_get_char(c);
         /* XXX handle zero width characters */
-        is_wide = g_unichar_iswide(uc);
+        is_wide = vt100_is_wide_char(uc);
         ctype = g_unichar_type(uc);
         /* XXX should this also include spacing marks? */
         is_combining = ctype == G_UNICODE_ENCLOSING_MARK
diff --git a/src/unicode-extra.c b/src/unicode-extra.c
new file mode 100644
index 0000000..3a59378
--- /dev/null
+++ b/src/unicode-extra.c
@@ -0,0 +1,95 @@
+#include <glib.h>
+
+#include "vt100.h"
+
+/*
+ * so, here's the story. unicode doesn't actually define monospace width for
+ * characters in a way that's useful. there's an "east asian width" property
+ * that mostly works, but leaves a bunch of things ambiguous (if you're
+ * displaying this as part of a bunch of east asian text, then it's wide, but
+ * if you're not, then it's narrow). we in general treat ambiguous characters
+ * as narrow for now (although this should perhaps be an option in the future).
+ * one place where this does not work out, though, is emoji. emoji do not have
+ * a useful width property (see
+ * http://www.unicode.org/L2/L2016/16027-emoji-terminals-eaw.pdf), and even the
+ * proposal in that link to make every character with Emoji_Presentation=true a
+ * wide character isn't really how things work in practice - for instance,
+ * U+231A (WATCH) is rendered by most monospace fonts (that I've seen, anyway)
+ * as a narrow character. as far as i can tell, it appears (although i'm not
+ * certain of this) that all BMP characters with Emoji_Presentation=true are
+ * narrow and all astral plane characters with Emoji_Presentation=true are
+ * wide, so that's what i'm going to go with here. character ranges and data in
+ * this file are taken from
+ * http://www.unicode.org/Public/emoji/2.0//emoji-data.txt.
+ */
+struct vt100_char_range {
+    gunichar start;
+    gunichar end;
+};
+
+static struct vt100_char_range vt100_wide_emoji[] = {
+    { 0x1F004, 0x1F004 },
+    { 0x1F0CF, 0x1F0CF },
+    { 0x1F18E, 0x1F18E },
+    { 0x1F191, 0x1F19A },
+    { 0x1F1E6, 0x1F1FF },
+    { 0x1F201, 0x1F201 },
+    { 0x1F21A, 0x1F21A },
+    { 0x1F22F, 0x1F22F },
+    { 0x1F232, 0x1F236 },
+    { 0x1F238, 0x1F23A },
+    { 0x1F250, 0x1F251 },
+    { 0x1F300, 0x1F320 },
+    { 0x1F32D, 0x1F335 },
+    { 0x1F337, 0x1F37C },
+    { 0x1F37E, 0x1F393 },
+    { 0x1F3A0, 0x1F3CA },
+    { 0x1F3CF, 0x1F3D3 },
+    { 0x1F3E0, 0x1F3F0 },
+    { 0x1F3F4, 0x1F3F4 },
+    { 0x1F3F8, 0x1F43E },
+    { 0x1F440, 0x1F440 },
+    { 0x1F442, 0x1F4FC },
+    { 0x1F4FF, 0x1F53D },
+    { 0x1F54B, 0x1F54E },
+    { 0x1F550, 0x1F567 },
+    { 0x1F595, 0x1F596 },
+    { 0x1F5FB, 0x1F64F },
+    { 0x1F680, 0x1F6C5 },
+    { 0x1F6CC, 0x1F6CC },
+    { 0x1F6D0, 0x1F6D0 },
+    { 0x1F6EB, 0x1F6EC },
+    { 0x1F910, 0x1F918 },
+    { 0x1F980, 0x1F984 },
+    { 0x1F9C0, 0x1F9C0 },
+};
+
+static int vt100_is_wide_emoji(gunichar codepoint);
+
+int vt100_is_wide_char(gunichar codepoint)
+{
+    return g_unichar_iswide(codepoint) || vt100_is_wide_emoji(codepoint);
+}
+
+static int vt100_is_wide_emoji(gunichar codepoint)
+{
+    static size_t ranges = sizeof(vt100_wide_emoji) / sizeof(struct vt100_char_range);
+    ssize_t low = 0, high = ranges - 1;
+
+    do {
+        ssize_t cur = (high + low) / 2;
+        struct vt100_char_range range = vt100_wide_emoji[cur];
+
+        if (codepoint < range.start) {
+            high = cur - 1;
+        }
+        else if (codepoint > range.end) {
+            low = cur + 1;
+        }
+        else {
+            return 1;
+        }
+    } while (low <= high);
+
+    return 0;
+}
diff --git a/src/unicode-extra.h b/src/unicode-extra.h
new file mode 100644
index 0000000..eaa3dce
--- /dev/null
+++ b/src/unicode-extra.h
@@ -0,0 +1,8 @@
+#ifndef _VT100_UNICODE_EXTRA_H
+#define _VT100_UNICODE_EXTRA_H
+
+#include <glib.h>
+
+int vt100_is_wide_char(gunichar codepoint);
+
+#endif
diff --git a/src/vt100.h b/src/vt100.h
index a3429ef..28e02b8 100644
--- a/src/vt100.h
+++ b/src/vt100.h
@@ -6,5 +6,6 @@ struct vt100_screen;
 typedef struct vt100_screen VT100Screen;
 
 #include "screen.h"
+#include "unicode-extra.h"
 
 #endif
-- 
cgit v1.2.3