From 8c6ba8a3444e4b8640e7fe04f565cef57c850432 Mon Sep 17 00:00:00 2001
From: Tyge Lovset <tylovset@gmail.com>
Date: Mon, 2 Jan 2023 09:00:47 +0100
Subject: Added selected unicode character classes.

---
 src/utf8code.c | 220 ++++++++++++++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 209 insertions(+), 11 deletions(-)

(limited to 'src/utf8code.c')

diff --git a/src/utf8code.c b/src/utf8code.c
index 5dfb7d30..6fe8515e 100644
--- a/src/utf8code.c
+++ b/src/utf8code.c
@@ -112,17 +112,29 @@ int utf8_icmp_sv(const csview s1, const csview s2) {
     return (int)(s1.size - s2.size);
 }
 
-bool utf8_isspace(uint32_t c) {
-    static uint16_t t[] = {0x20, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x85, 0xA0,
-                           0x1680, 0x2028, 0x2029, 0x202F, 0x205F, 0x3000};
-    for (size_t i=0; i<sizeof t/sizeof *t; ++i)
-        if (c == t[i]) return true;
-    return (c >= 0x2000) & (c <= 0x200A);
-}
+typedef struct {
+  uint16_t lo;
+  uint16_t hi;
+} URange16;
+
+typedef struct {
+  const char *name;
+  const URange16 *r16;
+  int nr16;
+} UGroup;
+
+static const UGroup unicode_groups[];
+static const int num_unicode_groups;
+static const int Lt_group;
 
-bool utf8_isdigit(uint32_t c) {
-    return ((c >= '0') & (c <= '9')) ||
-           ((c >= 0xFF10) & (c <= 0xFF19));
+bool utf8_isgroup(int group, uint32_t c) {
+    for (int j=0; j<unicode_groups[group].nr16; ++j) {
+        if (c < unicode_groups[group].r16[j].lo)
+            return false;
+        if (c <= unicode_groups[group].r16[j].hi)
+            return true;
+    }
+    return false;
 }
 
 bool utf8_isxdigit(uint32_t c) {
@@ -141,6 +153,192 @@ bool utf8_isalnum(uint32_t c) {
 
 bool utf8_isalpha(uint32_t c) {
     if (c < 128) return isalpha(c) != 0;
-    return utf8_islower(c) || utf8_isupper(c);
+    return utf8_islower(c) || utf8_isupper(c) || utf8_isgroup(Lt_group, c);
 }
+
+static const URange16 Cc_range16[] = { // Control
+    { 0, 31 },
+    { 127, 159 },
+};
+
+static const URange16 Lt_range16[] = { // Title case
+    { 453, 453 },
+    { 456, 456 },
+    { 459, 459 },
+    { 498, 498 },
+    { 8072, 8079 },
+    { 8088, 8095 },
+    { 8104, 8111 },
+    { 8124, 8124 },
+    { 8140, 8140 },
+    { 8188, 8188 },
+};
+
+static const URange16 Nd_range16[] = { // Decimal number
+    { 48, 57 },
+    { 1632, 1641 },
+    { 1776, 1785 },
+    { 1984, 1993 },
+    { 2406, 2415 },
+    { 2534, 2543 },
+    { 2662, 2671 },
+    { 2790, 2799 },
+    { 2918, 2927 },
+    { 3046, 3055 },
+    { 3174, 3183 },
+    { 3302, 3311 },
+    { 3430, 3439 },
+    { 3558, 3567 },
+    { 3664, 3673 },
+    { 3792, 3801 },
+    { 3872, 3881 },
+    { 4160, 4169 },
+    { 4240, 4249 },
+    { 6112, 6121 },
+    { 6160, 6169 },
+    { 6470, 6479 },
+    { 6608, 6617 },
+    { 6784, 6793 },
+    { 6800, 6809 },
+    { 6992, 7001 },
+    { 7088, 7097 },
+    { 7232, 7241 },
+    { 7248, 7257 },
+    { 42528, 42537 },
+    { 43216, 43225 },
+    { 43264, 43273 },
+    { 43472, 43481 },
+    { 43504, 43513 },
+    { 43600, 43609 },
+    { 44016, 44025 },
+    { 65296, 65305 },
+};
+
+static const URange16 Nl_range16[] = { // Number letter
+    { 5870, 5872 },
+    { 8544, 8578 },
+    { 8581, 8584 },
+    { 12295, 12295 },
+    { 12321, 12329 },
+    { 12344, 12346 },
+    { 42726, 42735 },
+};
+
+static const URange16 Pc_range16[] = { // Connector punctuation
+    { 95, 95 },
+    { 8255, 8256 },
+    { 8276, 8276 },
+    { 65075, 65076 },
+    { 65101, 65103 },
+    { 65343, 65343 },
+};
+
+static const URange16 Pd_range16[] = { // Dash punctuation
+    { 45, 45 },
+    { 1418, 1418 },
+    { 1470, 1470 },
+    { 5120, 5120 },
+    { 6150, 6150 },
+    { 8208, 8213 },
+    { 11799, 11799 },
+    { 11802, 11802 },
+    { 11834, 11835 },
+    { 11840, 11840 },
+    { 11869, 11869 },
+    { 12316, 12316 },
+    { 12336, 12336 },
+    { 12448, 12448 },
+    { 65073, 65074 },
+    { 65112, 65112 },
+    { 65123, 65123 },
+    { 65293, 65293 },
+};
+
+static const URange16 Pf_range16[] = { // Final punctuation
+    { 187, 187 },
+    { 8217, 8217 },
+    { 8221, 8221 },
+    { 8250, 8250 },
+    { 11779, 11779 },
+    { 11781, 11781 },
+    { 11786, 11786 },
+    { 11789, 11789 },
+    { 11805, 11805 },
+    { 11809, 11809 },
+};
+
+static const URange16 Pi_range16[] = { // Initial punctuation
+    { 171, 171 },
+    { 8216, 8216 },
+    { 8219, 8220 },
+    { 8223, 8223 },
+    { 8249, 8249 },
+    { 11778, 11778 },
+    { 11780, 11780 },
+    { 11785, 11785 },
+    { 11788, 11788 },
+    { 11804, 11804 },
+    { 11808, 11808 },
+};
+
+static const URange16 Sc_range16[] = { // Currency symbol
+    { 36, 36 },
+    { 162, 165 },
+    { 1423, 1423 },
+    { 1547, 1547 },
+    { 2046, 2047 },
+    { 2546, 2547 },
+    { 2555, 2555 },
+    { 2801, 2801 },
+    { 3065, 3065 },
+    { 3647, 3647 },
+    { 6107, 6107 },
+    { 8352, 8384 },
+    { 43064, 43064 },
+    { 65020, 65020 },
+    { 65129, 65129 },
+    { 65284, 65284 },
+    { 65504, 65505 },
+    { 65509, 65510 },
+};
+
+static const URange16 Zl_range16[] = { // Line separator
+    { 8232, 8232 },
+};
+
+static const URange16 Zp_range16[] = { // Paragraph separator
+    { 8233, 8233 },
+};
+
+static const URange16 Zs_range16[] = { // Space separator
+    { 32, 32 },
+    { 160, 160 },
+    { 5760, 5760 },
+    { 8192, 8202 },
+    { 8239, 8239 },
+    { 8287, 8287 },
+    { 12288, 12288 },
+};
+
+#define UNI_ENTRY(Code) \
+    { #Code, Code##_range16, sizeof(Code##_range16)/2 }
+
+static const UGroup unicode_groups[] = {
+    UNI_ENTRY(Cc),
+    UNI_ENTRY(Lt),
+    UNI_ENTRY(Nd),
+    UNI_ENTRY(Nl),
+    UNI_ENTRY(Pc),
+    UNI_ENTRY(Pd),
+    UNI_ENTRY(Pf),
+    UNI_ENTRY(Pi),
+    UNI_ENTRY(Sc),
+    UNI_ENTRY(Zl),
+    UNI_ENTRY(Zp),
+    UNI_ENTRY(Zs),
+};
+
+static const int Lt_group = 1;
+static const int num_unicode_groups = sizeof unicode_groups / sizeof unicode_groups[0];
+
 #endif
-- 
cgit v1.2.3