From f93b98a69d5da5dc75c27828c811279739dce47e Mon Sep 17 00:00:00 2001 From: "Reece H. Dunn" Date: Sat, 13 May 2017 21:18:54 +0100 Subject: [PATCH] Add Emoji support from emoji-data.txt. --- .gitignore | 1 + Makefile.am | 8 +- README.md | 3 +- src/include/ucd/ucd.h | 3 + src/proplist.c | 198 ++++++++++++++++++++++++++++++++++++++---- tools/printdata.py | 12 ++- tools/ucd.py | 4 + 7 files changed, 206 insertions(+), 23 deletions(-) diff --git a/.gitignore b/.gitignore index 058fa1c5..fdfc86f9 100644 --- a/.gitignore +++ b/.gitignore @@ -12,6 +12,7 @@ # build output: +data/emoji data/ucd src/libucd.la diff --git a/Makefile.am b/Makefile.am index 2b208ac8..fd7f3ed0 100644 --- a/Makefile.am +++ b/Makefile.am @@ -53,12 +53,17 @@ dist-hook: ChangeLog EXTRA_DIST += ChangeLog -############################# Unicode Character Database ###################### +############################# Unicode Data #################################### +EMOJI_VERSION=4.0 UCD_VERSION=@UCD_VERSION@ UCD_ROOTDIR=data/ucd UCD_SRCDIR=http://www.unicode.org/Public +data/emoji/emoji-data.txt: + mkdir -pv data/emoji + curl ${UCD_SRCDIR}/emoji/${EMOJI_VERSION}/emoji-data.txt > $@ + data/ucd/PropList.txt: mkdir -pv data/ucd curl ${UCD_SRCDIR}/${UCD_VERSION}/ucd/PropList.txt > $@ @@ -141,6 +146,7 @@ tests_printucddata_cpp_SOURCES = tests/printucddata_cpp.cpp tests_printucddata_cpp_LDADD = src/libucd.la tests/unicode-data.expected: tools/printdata.py tools/ucd.py \ + data/emoji/emoji-data.txt \ data/ucd/UnicodeData.txt \ data/ucd/PropList.txt \ data/ucd/DerivedCoreProperties.txt \ diff --git a/README.md b/README.md index 8a758b14..6a583766 100644 --- a/README.md +++ b/README.md @@ -17,7 +17,8 @@ this information within a C or C++ program. The project uses and supports the following sources of Unicode codepoint data: -* [Unicode Character Database](http://www.unicode.org/Public/9.0.0/ucd/) +* [Unicode Character Database](http://www.unicode.org/Public/9.0.0/ucd/) 9.0.0 +* [Unicode Emoji](http://www.unicode.org/Public/emoji/4.0/) 4.0 (UTR #51) * [ConScript Unicode Registry](http://www.evertype.com/standards/csur/) ## Build Dependencies diff --git a/src/include/ucd/ucd.h b/src/include/ucd/ucd.h index 82228594..486edce1 100644 --- a/src/include/ucd/ucd.h +++ b/src/include/ucd/ucd.h @@ -328,6 +328,7 @@ ucd_script ucd_lookup_script(codepoint_t c); /** @brief Properties */ typedef uint64_t ucd_property; + #define UCD_PROPERTY_WHITE_SPACE 0x0000000000000001ull /**< @brief White_Space */ #define UCD_PROPERTY_BIDI_CONTROL 0x0000000000000002ull /**< @brief Bidi_Control */ #define UCD_PROPERTY_JOIN_CONTROL 0x0000000000000004ull /**< @brief Join_Control */ @@ -361,6 +362,7 @@ typedef uint64_t ucd_property; #define UCD_PROPERTY_PATTERN_WHITE_SPACE 0x0000000040000000ull /**< @brief Pattern_White_Space */ #define UCD_PROPERTY_PATTERN_SYNTAX 0x0000000080000000ull /**< @brief Pattern_Syntax */ #define UCD_PROPERTY_PREPENDED_CONCATENATION_MARK 0x0000000100000000ull /**< @brief Prepended_Concatenation_Mark */ +#define UCD_PROPERTY_EMOJI 0x0000000200000000ull /**< @brief Emoji */ /** @brief Return the properties of the specified codepoint. * @@ -853,6 +855,7 @@ namespace ucd Pattern_White_Space = UCD_PROPERTY_PATTERN_WHITE_SPACE, /**< @brief Pattern_White_Space */ Pattern_Syntax = UCD_PROPERTY_PATTERN_SYNTAX, /**< @brief Pattern_Syntax */ Prepended_Concatenation_Mark = UCD_PROPERTY_PREPENDED_CONCATENATION_MARK, /**< @brief Prepended_Concatenation_Mark */ + Emoji = UCD_PROPERTY_EMOJI, /**< @brief Emoji */ }; /** @brief Return the properties of the specified codepoint. diff --git a/src/proplist.c b/src/proplist.c index 9cf15f83..73c4c9d4 100644 --- a/src/proplist.c +++ b/src/proplist.c @@ -163,6 +163,7 @@ static ucd_property properties_Ll(codepoint_t c) if (c >= 0x210A && c <= 0x2113) return UCD_PROPERTY_OTHER_MATH; // Ll|Lu if (c >= 0x212F && c <= 0x2131) return UCD_PROPERTY_OTHER_MATH; // Ll|Lu if (c >= 0x2133 && c <= 0x2134) return UCD_PROPERTY_OTHER_MATH; // Ll|Lu + if (c == 0x2139) return UCD_PROPERTY_EMOJI; if (c >= 0x213C && c <= 0x213F) return UCD_PROPERTY_OTHER_MATH; // Ll|Lu if (c >= 0x2145 && c <= 0x2147) return UCD_PROPERTY_OTHER_MATH; // Ll|Lu if (c >= 0x2148 && c <= 0x2149) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_SOFT_DOTTED; @@ -1157,7 +1158,7 @@ static ucd_property properties_Nd(codepoint_t c) switch (c & 0xFFFFFF00) { case 0x0000: - if (c >= 0x0030 && c <= 0x0039) return UCD_PROPERTY_HEX_DIGIT | UCD_PROPERTY_ASCII_HEX_DIGIT; + if (c >= 0x0030 && c <= 0x0039) return UCD_PROPERTY_HEX_DIGIT | UCD_PROPERTY_ASCII_HEX_DIGIT | UCD_PROPERTY_EMOJI; break; case 0xFF00: if (c >= 0xFF10 && c <= 0xFF19) return UCD_PROPERTY_HEX_DIGIT; @@ -1231,7 +1232,7 @@ static ucd_property properties_Pd(codepoint_t c) return UCD_PROPERTY_DASH | UCD_PROPERTY_PATTERN_SYNTAX; case 0x3000: if (c == 0x301C) return UCD_PROPERTY_DASH | UCD_PROPERTY_PATTERN_SYNTAX; - if (c == 0x3030) return UCD_PROPERTY_DASH | UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x3030) return UCD_PROPERTY_DASH | UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; break; case 0xFE00: if (c == 0xFE63) return UCD_PROPERTY_DASH | UCD_PROPERTY_HYPHEN | UCD_PROPERTY_OTHER_MATH; @@ -1319,7 +1320,9 @@ static ucd_property properties_Po(codepoint_t c) case 0x0000: if (c == 0x0021) return UCD_PROPERTY_TERMINAL_PUNCTUATION | UCD_PROPERTY_SENTENCE_TERMINAL | UCD_PROPERTY_PATTERN_SYNTAX; if (c == 0x0022) return UCD_PROPERTY_QUOTATION_MARK | UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x0023) return UCD_PROPERTY_EMOJI | UCD_PROPERTY_PATTERN_SYNTAX; if (c == 0x0027) return UCD_PROPERTY_QUOTATION_MARK | UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x002A) return UCD_PROPERTY_EMOJI | UCD_PROPERTY_PATTERN_SYNTAX; if (c == 0x002C) return UCD_PROPERTY_TERMINAL_PUNCTUATION | UCD_PROPERTY_PATTERN_SYNTAX; if (c == 0x002E) return UCD_PROPERTY_TERMINAL_PUNCTUATION | UCD_PROPERTY_SENTENCE_TERMINAL | UCD_PROPERTY_PATTERN_SYNTAX; if (c >= 0x003A && c <= 0x003B) return UCD_PROPERTY_TERMINAL_PUNCTUATION | UCD_PROPERTY_PATTERN_SYNTAX; @@ -1410,10 +1413,12 @@ static ucd_property properties_Po(codepoint_t c) if (c >= 0x2020 && c <= 0x2027) return UCD_PROPERTY_PATTERN_SYNTAX; if (c >= 0x2032 && c <= 0x2034) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_OTHER_MATH; if (c >= 0x2030 && c <= 0x2038) return UCD_PROPERTY_PATTERN_SYNTAX; - if (c >= 0x203C && c <= 0x203D) return UCD_PROPERTY_TERMINAL_PUNCTUATION | UCD_PROPERTY_SENTENCE_TERMINAL | UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x203C) return UCD_PROPERTY_TERMINAL_PUNCTUATION | UCD_PROPERTY_SENTENCE_TERMINAL | UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x203D) return UCD_PROPERTY_TERMINAL_PUNCTUATION | UCD_PROPERTY_SENTENCE_TERMINAL | UCD_PROPERTY_PATTERN_SYNTAX; if (c >= 0x203B && c <= 0x203E) return UCD_PROPERTY_PATTERN_SYNTAX; if (c >= 0x2041 && c <= 0x2043) return UCD_PROPERTY_PATTERN_SYNTAX; - if (c >= 0x2047 && c <= 0x2049) return UCD_PROPERTY_TERMINAL_PUNCTUATION | UCD_PROPERTY_SENTENCE_TERMINAL | UCD_PROPERTY_PATTERN_SYNTAX; + if (c >= 0x2047 && c <= 0x2048) return UCD_PROPERTY_TERMINAL_PUNCTUATION | UCD_PROPERTY_SENTENCE_TERMINAL | UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x2049) return UCD_PROPERTY_TERMINAL_PUNCTUATION | UCD_PROPERTY_SENTENCE_TERMINAL | UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; if (c >= 0x204A && c <= 0x2051) return UCD_PROPERTY_PATTERN_SYNTAX; if (c == 0x2053) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_DASH; if (c >= 0x2055 && c <= 0x205E) return UCD_PROPERTY_PATTERN_SYNTAX; @@ -1427,6 +1432,7 @@ static ucd_property properties_Po(codepoint_t c) if (c == 0x3001) return UCD_PROPERTY_TERMINAL_PUNCTUATION | UCD_PROPERTY_PATTERN_SYNTAX; if (c == 0x3002) return UCD_PROPERTY_TERMINAL_PUNCTUATION | UCD_PROPERTY_SENTENCE_TERMINAL | UCD_PROPERTY_PATTERN_SYNTAX; if (c == 0x3003) return UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x303D) return UCD_PROPERTY_EMOJI; if (c == 0x30FB) return UCD_PROPERTY_HYPHEN; break; case 0xA400: @@ -1660,6 +1666,9 @@ static ucd_property properties_Sk(codepoint_t c) if (c == 0xFF40) return UCD_PROPERTY_DIACRITIC; if (c == 0xFFE3) return UCD_PROPERTY_DIACRITIC; break; + case 0x01F300: + if (c >= 0x01F3FB && c <= 0x01F3FF) return UCD_PROPERTY_EMOJI; + break; } return 0; } @@ -1678,16 +1687,23 @@ static ucd_property properties_Sm(codepoint_t c) break; case 0x2100: if (c == 0x2118) return UCD_PROPERTY_OTHER_ID_START; + if (c == 0x2194) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; if (c >= 0x2190) return UCD_PROPERTY_PATTERN_SYNTAX; break; case 0x2200: if (c == 0x2212) return UCD_PROPERTY_DASH | UCD_PROPERTY_PATTERN_SYNTAX; return UCD_PROPERTY_PATTERN_SYNTAX; case 0x2300: + return UCD_PROPERTY_PATTERN_SYNTAX; case 0x2500: + if (c >= 0x25FB && c <= 0x25FE) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + return UCD_PROPERTY_PATTERN_SYNTAX; case 0x2600: case 0x2700: + return UCD_PROPERTY_PATTERN_SYNTAX; case 0x2900: + if (c >= 0x2934 && c <= 0x2935) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + return UCD_PROPERTY_PATTERN_SYNTAX; case 0x2A00: case 0x2B00: return UCD_PROPERTY_PATTERN_SYNTAX; @@ -1700,16 +1716,18 @@ static ucd_property properties_So(codepoint_t c) switch (c & 0xFFFFFF00) { case 0x0000: + if (c == 0x00A9) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x00AE) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; return UCD_PROPERTY_PATTERN_SYNTAX; case 0x2100: + if (c == 0x2122) return UCD_PROPERTY_EMOJI; if (c == 0x2129) return UCD_PROPERTY_OTHER_MATH; if (c == 0x212E) return UCD_PROPERTY_OTHER_ID_START; - if (c >= 0x2195 && c <= 0x2199) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; - if (c >= 0x219C && c <= 0x219F) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; - if (c >= 0x21A1 && c <= 0x21A2) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; - if (c >= 0x21A4 && c <= 0x21A5) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; if (c == 0x21A8) return UCD_PROPERTY_PATTERN_SYNTAX; - if (c >= 0x21A7 && c <= 0x21AD) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x21A9) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x21AA) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2195 && c <= 0x2199) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x219C && c <= 0x21AD) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; if (c >= 0x21B0 && c <= 0x21B1) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; if (c >= 0x21B6 && c <= 0x21B7) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; if (c >= 0x21AF && c <= 0x21BB) return UCD_PROPERTY_PATTERN_SYNTAX; @@ -1722,28 +1740,37 @@ static ucd_property properties_So(codepoint_t c) if (c >= 0x21D5 && c <= 0x21F3) return UCD_PROPERTY_PATTERN_SYNTAX; break; case 0x2300: - if (c >= 0x2300 && c <= 0x2307) return UCD_PROPERTY_PATTERN_SYNTAX; + if (c <= 0x2307) return UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x231A) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x231B) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; if (c >= 0x230C && c <= 0x231F) return UCD_PROPERTY_PATTERN_SYNTAX; - if (c >= 0x2322 && c <= 0x2328) return UCD_PROPERTY_PATTERN_SYNTAX; + if (c >= 0x2322 && c <= 0x2327) return UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x2328) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; if (c >= 0x232B && c <= 0x237B) return UCD_PROPERTY_PATTERN_SYNTAX; if (c >= 0x237D && c <= 0x239A) return UCD_PROPERTY_PATTERN_SYNTAX; if (c >= 0x23B4 && c <= 0x23B5) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; if (c == 0x23B7) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x23CF) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; if (c == 0x23D0) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; if (c >= 0x23B4 && c <= 0x23DB) return UCD_PROPERTY_PATTERN_SYNTAX; if (c == 0x23E2) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; + if (c >= 0x23E9 && c <= 0x23F3) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x23F8 && c <= 0x23FA) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; if (c >= 0x23E3 && c <= 0x23FE) return UCD_PROPERTY_PATTERN_SYNTAX; break; case 0x2400: - if (c >= 0x2400 && c <= 0x2426) return UCD_PROPERTY_PATTERN_SYNTAX; - if (c >= 0x2440 && c <= 0x244A) return UCD_PROPERTY_PATTERN_SYNTAX; + if (c >= 0x2400 && c <= 0x244A) return UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x24C2) return UCD_PROPERTY_OTHER_ALPHABETIC | UCD_PROPERTY_OTHER_UPPERCASE | UCD_PROPERTY_EMOJI; if (c >= 0x24B6 && c <= 0x24CF) return UCD_PROPERTY_OTHER_ALPHABETIC | UCD_PROPERTY_OTHER_UPPERCASE; if (c >= 0x24D0 && c <= 0x24E9) return UCD_PROPERTY_OTHER_ALPHABETIC | UCD_PROPERTY_OTHER_LOWERCASE; break; case 0x2500: if (c >= 0x25A0 && c <= 0x25A1) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; - if (c >= 0x25AE && c <= 0x25B6) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; - if (c >= 0x25BC && c <= 0x25C0) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; + if (c >= 0x25AA && c <= 0x25AB) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x25AE && c <= 0x25B5) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x25B6) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x25BC && c <= 0x25BF) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x25C0) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; if (c >= 0x25C6 && c <= 0x25C7) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; if (c >= 0x25CA && c <= 0x25CB) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; if (c >= 0x25CF && c <= 0x25D3) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; @@ -1752,22 +1779,85 @@ static ucd_property properties_So(codepoint_t c) if (c >= 0x25E7 && c <= 0x25EC) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; return UCD_PROPERTY_PATTERN_SYNTAX; case 0x2600: + if (c <= 0x2604) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; if (c >= 0x2605 && c <= 0x2606) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; - if (c == 0x2640) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; - if (c == 0x2642) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; - if (c >= 0x2660 && c <= 0x2663) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x260E) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2611) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2614 && c <= 0x2615) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2618) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x261D) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2620) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2622 && c <= 0x2623) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2626) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x262A) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x262E && c <= 0x262F) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2638 && c <= 0x263A) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2640) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2642) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2648 && c <= 0x2653) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2660) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2661 && c <= 0x2662) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x2663) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2665 && c <= 0x2666) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2668) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; if (c >= 0x266D && c <= 0x266E) return UCD_PROPERTY_OTHER_MATH | UCD_PROPERTY_PATTERN_SYNTAX; + if (c == 0x267B) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x267F) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2692 && c <= 0x2697) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2699) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x269B && c <= 0x269C) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x26A0 && c <= 0x26A1) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x26AA && c <= 0x26AB) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x26B0 && c <= 0x26B1) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x26BD && c <= 0x26BE) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x26C4 && c <= 0x26C5) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x26C8) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x26CE && c <= 0x26CF) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x26D1) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x26D3 && c <= 0x26D4) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x26E9 && c <= 0x26EA) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x26F0 && c <= 0x26F5) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x26F7 && c <= 0x26FA) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x26FD) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; return UCD_PROPERTY_PATTERN_SYNTAX; case 0x2700: + if (c == 0x2702) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2705) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2708 && c <= 0x270D) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x270F) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2712) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2714) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2716) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x271D) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2721) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2728) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2733 && c <= 0x2734) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2744) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2747) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x274C) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x274E) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2753 && c <= 0x2755) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2757) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2763 && c <= 0x2764) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2795 && c <= 0x2797) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x27A1) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x27B0) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x27BF) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + return UCD_PROPERTY_PATTERN_SYNTAX; case 0x2800: + return UCD_PROPERTY_PATTERN_SYNTAX; case 0x2B00: + if (c >= 0x2B05 && c <= 0x2B07) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c >= 0x2B1B && c <= 0x2B1C) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2B50) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; + if (c == 0x2B55) return UCD_PROPERTY_PATTERN_SYNTAX | UCD_PROPERTY_EMOJI; return UCD_PROPERTY_PATTERN_SYNTAX; case 0x2E00: if (c >= 0x2E80 && c <= 0x2E99) return UCD_PROPERTY_RADICAL; if (c >= 0x2E9B && c <= 0x2EF3) return UCD_PROPERTY_RADICAL; break; case 0x2F00: - if (c >= 0x2F00 && c <= 0x2FD5) return UCD_PROPERTY_RADICAL; + if (c <= 0x2FD5) return UCD_PROPERTY_RADICAL; if (c >= 0x2FF0 && c <= 0x2FF1) return UCD_PROPERTY_IDS_BINARY_OPERATOR; if (c >= 0x2FF2 && c <= 0x2FF3) return UCD_PROPERTY_IDS_TRINARY_OPERATOR; if (c >= 0x2FF4 && c <= 0x2FFB) return UCD_PROPERTY_IDS_BINARY_OPERATOR; @@ -1776,11 +1866,81 @@ static ucd_property properties_So(codepoint_t c) if (c >= 0x3012 && c <= 0x3013) return UCD_PROPERTY_PATTERN_SYNTAX; if (c == 0x3020) return UCD_PROPERTY_PATTERN_SYNTAX; break; + case 0x3200: + if (c == 0x3297) return UCD_PROPERTY_EMOJI; + if (c == 0x3299) return UCD_PROPERTY_EMOJI; + break; + case 0x01F000: + if (c == 0x01F004) return UCD_PROPERTY_EMOJI; + if (c == 0x01F0CF) return UCD_PROPERTY_EMOJI; + break; case 0x01F100: if (c >= 0x01F130 && c <= 0x01F149) return UCD_PROPERTY_OTHER_ALPHABETIC | UCD_PROPERTY_OTHER_UPPERCASE; if (c >= 0x01F150 && c <= 0x01F169) return UCD_PROPERTY_OTHER_ALPHABETIC | UCD_PROPERTY_OTHER_UPPERCASE; + if (c >= 0x01F170 && c <= 0x01F171) return UCD_PROPERTY_OTHER_ALPHABETIC | UCD_PROPERTY_OTHER_UPPERCASE | UCD_PROPERTY_EMOJI; + if (c >= 0x01F17E && c <= 0x01F17F) return UCD_PROPERTY_OTHER_ALPHABETIC | UCD_PROPERTY_OTHER_UPPERCASE | UCD_PROPERTY_EMOJI; if (c >= 0x01F170 && c <= 0x01F189) return UCD_PROPERTY_OTHER_ALPHABETIC | UCD_PROPERTY_OTHER_UPPERCASE; + if (c == 0x01F18E) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F191 && c <= 0x01F19A) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F1E6) return UCD_PROPERTY_EMOJI; break; + case 0x01F200: + if (c >= 0x01F201 && c <= 0x01F202) return UCD_PROPERTY_EMOJI; + if (c == 0x01F21A) return UCD_PROPERTY_EMOJI; + if (c == 0x01F22F) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F232 && c <= 0x01F23A) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F250 && c <= 0x01F251) return UCD_PROPERTY_EMOJI; + break; + case 0x01F300: + if (c <= 0x01F321) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F324 && c <= 0x01F393) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F396 && c <= 0x01F397) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F399 && c <= 0x01F39B) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F39E && c <= 0x01F3F0) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F3F3 && c <= 0x01F3F5) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F3F7 && c <= 0x01F3FA) return UCD_PROPERTY_EMOJI; + break; + case 0x01F400: + if (c != 0x01F4FE) return UCD_PROPERTY_EMOJI; + break; + case 0x01F500: + if (c <= 0x01F53D) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F549 && c <= 0x01F54E) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F550 && c <= 0x01F567) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F56F && c <= 0x01F570) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F573 && c <= 0x01F57A) return UCD_PROPERTY_EMOJI; + if (c == 0x01F587) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F58A && c <= 0x01F58D) return UCD_PROPERTY_EMOJI; + if (c == 0x01F590) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F595 && c <= 0x01F596) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F5A4 && c <= 0x01F5A5) return UCD_PROPERTY_EMOJI; + if (c == 0x01F5A8) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F5B1 && c <= 0x01F5B2) return UCD_PROPERTY_EMOJI; + if (c == 0x01F5BC) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F5C2 && c <= 0x01F5C4) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F5D1 && c <= 0x01F5D3) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F5DC && c <= 0x01F5DE) return UCD_PROPERTY_EMOJI; + if (c == 0x01F5E1) return UCD_PROPERTY_EMOJI; + if (c == 0x01F5E3) return UCD_PROPERTY_EMOJI; + if (c == 0x01F5E8) return UCD_PROPERTY_EMOJI; + if (c == 0x01F5EF) return UCD_PROPERTY_EMOJI; + if (c == 0x01F5F3) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F5FA) return UCD_PROPERTY_EMOJI; + break; + case 0x01F600: + if (c <= 0x01F64F) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F680 && c <= 0x01F6C5) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F6CB && c <= 0x01F6D2) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F6E0 && c <= 0x01F6E5) return UCD_PROPERTY_EMOJI; + if (c == 0x01F6E9) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F6EB && c <= 0x01F6EC) return UCD_PROPERTY_EMOJI; + if (c == 0x01F6F0) return UCD_PROPERTY_EMOJI; + if (c >= 0x01F6F3 && c <= 0x01F6F6) return UCD_PROPERTY_EMOJI; + break; + case 0x01F900: + if (c == 0x01F93B) return 0; + if (c == 0x01F946) return 0; + return UCD_PROPERTY_EMOJI; } return 0; } diff --git a/tools/printdata.py b/tools/printdata.py index 40ee24fd..3d904f63 100755 --- a/tools/printdata.py +++ b/tools/printdata.py @@ -22,16 +22,23 @@ import sys import ucd ucd_rootdir = sys.argv[1] +emoji_rootdir = 'data/emoji' csur_rootdir = 'data/csur' null = ucd.CodePoint('0000') +properties = [ + (ucd_rootdir, 'PropList'), + (ucd_rootdir, 'DerivedCoreProperties'), + (emoji_rootdir, 'emoji-data') +] + unicode_chars = {} for data in ucd.parse_ucd_data(ucd_rootdir, 'UnicodeData'): for codepoint in data['CodePoint']: unicode_chars[codepoint] = data -for propfile in ['PropList', 'DerivedCoreProperties']: - for data in ucd.parse_ucd_data(ucd_rootdir, propfile): +for propdir, propfile in properties: + for data in ucd.parse_ucd_data(propdir, propfile): for codepoint in data['Range']: try: unicode_chars[codepoint][data['Property']] = 1 @@ -154,6 +161,7 @@ def properties(data): props += (2 ** 30) * data.get('Pattern_White_Space', 0) props += (2 ** 31) * data.get('Pattern_Syntax', 0) props += (2 ** 32) * data.get('Prepended_Concatenation_Mark', 0) + props += (2 ** 33) * data.get('Emoji', 0) # emoji-data return props if __name__ == '__main__': diff --git a/tools/ucd.py b/tools/ucd.py index 361cebd9..466997dc 100755 --- a/tools/ucd.py +++ b/tools/ucd.py @@ -105,6 +105,10 @@ def strlist(x): data_items = { # Unicode Character Data: + 'emoji-data': [ + ('Range', codepoint), + ('Property', string) + ], 'Blocks': [ ('Range', codepoint), ('Name', string)