Add to codepoint data whether the codepoint is a character or not

Signed-off-by: Mattias Andrée <m@maandree.se>

2 files changed, 17 insertions, 2 deletions

diff --git a/UnicodeData.parse.c b/UnicodeData.parse.c
index a5879b9..f9b676c 100644
--- a/UnicodeData.parse.c
+++ b/UnicodeData.parse.c
@@ -129,10 +129,12 @@ static int
 output_character_data(char *text)
 {
 	char *field, *numeric_value = NULL, *simple_uppercase_mapping;
+	unsigned long int codepoint;
 	const char *formatted;
 	size_t n, i;
 
-	if (!(field = split(&text))) return 0; /* codepoint */
+	if (!(field = split(&text))) return 0;
+	codepoint = strtoul(field, &field, 16);
 
 	if (!(field = split(&text))) return 0;
 	add_linef(".name = \"%s\"", field);
@@ -145,6 +147,13 @@ output_character_data(char *text)
 
 	if (!(field = split(&text))) return 0;
 	add_linef(", .bidirectional_category = BIDIRECTIONAL_CATEGORY_%s", field);
+	if (strcmp(field, "Cn")) {
+		add_linef(", .noncharacter = 0");
+	} else if ((0xFDD0u <= codepoint && codepoint <= 0xFDEFu) || (codepoint & 0xFFFEu) == 0xFFFEu) {
+		add_linef(", .noncharacter = 1");
+	} else {
+		add_linef(", .noncharacter = 0");
+	}
 
 	if (!(field = split(&text))) return 0;
 	if (!*field) {
diff --git a/libcmap.h b/libcmap.h
index 50d2146..0f765a1 100644
--- a/libcmap.h
+++ b/libcmap.h
@@ -76,6 +76,11 @@ enum libcmap_general_category {
 	LIBCMAP_GENERAL_CATEGORY_OTHER_FORMAT,
 	LIBCMAP_GENERAL_CATEGORY_OTHER_SURROGATE,
 	LIBCMAP_GENERAL_CATEGORY_OTHER_PRIVATE_USE,
+
+	/**
+	 * This category has two subcategories: noncharacter and reserved
+	 * If .noncharacter it is not a character, otherwise it is reserved
+	 */
 	LIBCMAP_GENERAL_CATEGORY_OTHER_NOT_ASSIGNED
 };
 
@@ -240,7 +245,8 @@ struct libcmap_codepoint_data {
 	uint32_t simple_titlecase_mapping : 21;
 	uint32_t decomposition_mapping_size : 6;
 	uint32_t bidirectional_mirrored : 1;
-	uint32_t : 4;
+	uint32_t noncharacter : 1;
+	uint32_t : 3;
 };