| /* Unit tests for utilities |
| * Copyright (C) 2010 Red Hat, Inc. |
| * Copyright (C) 2011 Google, Inc. |
| * |
| * This work is provided "as is"; redistribution and modification |
| * in whole or in part, in any medium, physical or electronic is |
| * permitted without restriction. |
| * |
| * This work is distributed in the hope that it will be useful, |
| * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. |
| * |
| * In no event shall the authors or contributors be liable for any |
| * direct, indirect, incidental, special, exemplary, or consequential |
| * damages (including, but not limited to, procurement of substitute |
| * goods or services; loss of use, data, or profits; or business |
| * interruption) however caused and on any theory of liability, whether |
| * in contract, strict liability, or tort (including negligence or |
| * otherwise) arising in any way out of the use of this software, even |
| * if advised of the possibility of such damage. |
| * |
| * Author: Matthias Clasen, Behdad Esfahbod |
| */ |
| |
| /* We are testing some deprecated APIs here */ |
| #define GLIB_DISABLE_DEPRECATION_WARNINGS |
| |
| #include "glib.h" |
| |
| static void |
| test_unichar_validate (void) |
| { |
| g_assert (g_unichar_validate ('j')); |
| g_assert (g_unichar_validate (8356)); |
| g_assert (g_unichar_validate (8356)); |
| g_assert (g_unichar_validate (0xfdd1)); |
| g_assert (g_unichar_validate (917760)); |
| g_assert (!g_unichar_validate (0x110000)); |
| } |
| |
| static void |
| test_unichar_character_type (void) |
| { |
| guint i; |
| struct { |
| GUnicodeType type; |
| gunichar c; |
| } examples[] = { |
| { G_UNICODE_CONTROL, 0x000D }, |
| { G_UNICODE_FORMAT, 0x200E }, |
| /* G_UNICODE_UNASSIGNED */ |
| { G_UNICODE_PRIVATE_USE, 0xE000 }, |
| { G_UNICODE_SURROGATE, 0xD800 }, |
| { G_UNICODE_LOWERCASE_LETTER, 0x0061 }, |
| { G_UNICODE_MODIFIER_LETTER, 0x02B0 }, |
| { G_UNICODE_OTHER_LETTER, 0x3400 }, |
| { G_UNICODE_TITLECASE_LETTER, 0x01C5 }, |
| { G_UNICODE_UPPERCASE_LETTER, 0xFF21 }, |
| { G_UNICODE_COMBINING_MARK, 0x0903 }, |
| { G_UNICODE_ENCLOSING_MARK, 0x20DD }, |
| { G_UNICODE_NON_SPACING_MARK, 0xA806 }, |
| { G_UNICODE_DECIMAL_NUMBER, 0xFF10 }, |
| { G_UNICODE_LETTER_NUMBER, 0x16EE }, |
| { G_UNICODE_OTHER_NUMBER, 0x17F0 }, |
| { G_UNICODE_CONNECT_PUNCTUATION, 0x005F }, |
| { G_UNICODE_DASH_PUNCTUATION, 0x058A }, |
| { G_UNICODE_CLOSE_PUNCTUATION, 0x0F3B }, |
| { G_UNICODE_FINAL_PUNCTUATION, 0x2019 }, |
| { G_UNICODE_INITIAL_PUNCTUATION, 0x2018 }, |
| { G_UNICODE_OTHER_PUNCTUATION, 0x2016 }, |
| { G_UNICODE_OPEN_PUNCTUATION, 0x0F3A }, |
| { G_UNICODE_CURRENCY_SYMBOL, 0x20A0 }, |
| { G_UNICODE_MODIFIER_SYMBOL, 0x309B }, |
| { G_UNICODE_MATH_SYMBOL, 0xFB29 }, |
| { G_UNICODE_OTHER_SYMBOL, 0x00A6 }, |
| { G_UNICODE_LINE_SEPARATOR, 0x2028 }, |
| { G_UNICODE_PARAGRAPH_SEPARATOR, 0x2029 }, |
| { G_UNICODE_SPACE_SEPARATOR, 0x202F }, |
| }; |
| |
| for (i = 0; i < G_N_ELEMENTS (examples); i++) |
| { |
| g_assert_cmpint (g_unichar_type (examples[i].c), ==, examples[i].type); |
| } |
| } |
| |
| static void |
| test_unichar_break_type (void) |
| { |
| guint i; |
| struct { |
| GUnicodeBreakType type; |
| gunichar c; |
| } examples[] = { |
| { G_UNICODE_BREAK_MANDATORY, 0x2028 }, |
| { G_UNICODE_BREAK_CARRIAGE_RETURN, 0x000D }, |
| { G_UNICODE_BREAK_LINE_FEED, 0x000A }, |
| { G_UNICODE_BREAK_COMBINING_MARK, 0x0300 }, |
| { G_UNICODE_BREAK_SURROGATE, 0xD800 }, |
| { G_UNICODE_BREAK_ZERO_WIDTH_SPACE, 0x200B }, |
| { G_UNICODE_BREAK_INSEPARABLE, 0x2024 }, |
| { G_UNICODE_BREAK_NON_BREAKING_GLUE, 0x00A0 }, |
| { G_UNICODE_BREAK_CONTINGENT, 0xFFFC }, |
| { G_UNICODE_BREAK_SPACE, 0x0020 }, |
| { G_UNICODE_BREAK_AFTER, 0x05BE }, |
| { G_UNICODE_BREAK_BEFORE, 0x02C8 }, |
| { G_UNICODE_BREAK_BEFORE_AND_AFTER, 0x2014 }, |
| { G_UNICODE_BREAK_HYPHEN, 0x002D }, |
| { G_UNICODE_BREAK_NON_STARTER, 0x17D6 }, |
| { G_UNICODE_BREAK_OPEN_PUNCTUATION, 0x0028 }, |
| { G_UNICODE_BREAK_CLOSE_PARANTHESIS, 0x0029 }, |
| { G_UNICODE_BREAK_CLOSE_PUNCTUATION, 0x007D }, |
| { G_UNICODE_BREAK_QUOTATION, 0x0022 }, |
| { G_UNICODE_BREAK_EXCLAMATION, 0x0021 }, |
| { G_UNICODE_BREAK_IDEOGRAPHIC, 0x2E80 }, |
| { G_UNICODE_BREAK_NUMERIC, 0x0030 }, |
| { G_UNICODE_BREAK_INFIX_SEPARATOR, 0x002C }, |
| { G_UNICODE_BREAK_SYMBOL, 0x002F }, |
| { G_UNICODE_BREAK_ALPHABETIC, 0x0023 }, |
| { G_UNICODE_BREAK_PREFIX, 0x0024 }, |
| { G_UNICODE_BREAK_POSTFIX, 0x0025 }, |
| { G_UNICODE_BREAK_COMPLEX_CONTEXT, 0x0E01 }, |
| { G_UNICODE_BREAK_AMBIGUOUS, 0x00F7 }, |
| { G_UNICODE_BREAK_UNKNOWN, 0xE000 }, |
| { G_UNICODE_BREAK_NEXT_LINE, 0x0085 }, |
| { G_UNICODE_BREAK_WORD_JOINER, 0x2060 }, |
| { G_UNICODE_BREAK_HANGUL_L_JAMO, 0x1100 }, |
| { G_UNICODE_BREAK_HANGUL_V_JAMO, 0x1160 }, |
| { G_UNICODE_BREAK_HANGUL_T_JAMO, 0x11A8 }, |
| { G_UNICODE_BREAK_HANGUL_LV_SYLLABLE, 0xAC00 }, |
| { G_UNICODE_BREAK_HANGUL_LVT_SYLLABLE, 0xAC01 }, |
| { G_UNICODE_BREAK_CONDITIONAL_JAPANESE_STARTER, 0x3041 }, |
| { G_UNICODE_BREAK_HEBREW_LETTER, 0x05D0 }, |
| { G_UNICODE_BREAK_REGIONAL_INDICATOR, 0x1F1F6 }, |
| { G_UNICODE_BREAK_EMOJI_BASE, 0x1F466 }, |
| { G_UNICODE_BREAK_EMOJI_MODIFIER, 0x1F3FB }, |
| { G_UNICODE_BREAK_ZERO_WIDTH_JOINER, 0x200D }, |
| }; |
| |
| for (i = 0; i < G_N_ELEMENTS (examples); i++) |
| { |
| g_assert_cmpint (g_unichar_break_type (examples[i].c), ==, examples[i].type); |
| } |
| } |
| |
| static void |
| test_unichar_script (void) |
| { |
| guint i; |
| struct { |
| GUnicodeScript script; |
| gunichar c; |
| } examples[] = { |
| { G_UNICODE_SCRIPT_COMMON, 0x002A }, |
| { G_UNICODE_SCRIPT_INHERITED, 0x1CED }, |
| { G_UNICODE_SCRIPT_INHERITED, 0x0670 }, |
| { G_UNICODE_SCRIPT_ARABIC, 0x060D }, |
| { G_UNICODE_SCRIPT_ARMENIAN, 0x0559 }, |
| { G_UNICODE_SCRIPT_BENGALI, 0x09CD }, |
| { G_UNICODE_SCRIPT_BOPOMOFO, 0x31B6 }, |
| { G_UNICODE_SCRIPT_CHEROKEE, 0x13A2 }, |
| { G_UNICODE_SCRIPT_COPTIC, 0x2CFD }, |
| { G_UNICODE_SCRIPT_CYRILLIC, 0x0482 }, |
| { G_UNICODE_SCRIPT_DESERET, 0x10401 }, |
| { G_UNICODE_SCRIPT_DEVANAGARI, 0x094D }, |
| { G_UNICODE_SCRIPT_ETHIOPIC, 0x1258 }, |
| { G_UNICODE_SCRIPT_GEORGIAN, 0x10FC }, |
| { G_UNICODE_SCRIPT_GOTHIC, 0x10341 }, |
| { G_UNICODE_SCRIPT_GREEK, 0x0375 }, |
| { G_UNICODE_SCRIPT_GUJARATI, 0x0A83 }, |
| { G_UNICODE_SCRIPT_GURMUKHI, 0x0A3C }, |
| { G_UNICODE_SCRIPT_HAN, 0x3005 }, |
| { G_UNICODE_SCRIPT_HANGUL, 0x1100 }, |
| { G_UNICODE_SCRIPT_HEBREW, 0x05BF }, |
| { G_UNICODE_SCRIPT_HIRAGANA, 0x309F }, |
| { G_UNICODE_SCRIPT_KANNADA, 0x0CBC }, |
| { G_UNICODE_SCRIPT_KATAKANA, 0x30FF }, |
| { G_UNICODE_SCRIPT_KHMER, 0x17DD }, |
| { G_UNICODE_SCRIPT_LAO, 0x0EDD }, |
| { G_UNICODE_SCRIPT_LATIN, 0x0061 }, |
| { G_UNICODE_SCRIPT_MALAYALAM, 0x0D3D }, |
| { G_UNICODE_SCRIPT_MONGOLIAN, 0x1843 }, |
| { G_UNICODE_SCRIPT_MYANMAR, 0x1031 }, |
| { G_UNICODE_SCRIPT_OGHAM, 0x169C }, |
| { G_UNICODE_SCRIPT_OLD_ITALIC, 0x10322 }, |
| { G_UNICODE_SCRIPT_ORIYA, 0x0B3C }, |
| { G_UNICODE_SCRIPT_RUNIC, 0x16EF }, |
| { G_UNICODE_SCRIPT_SINHALA, 0x0DBD }, |
| { G_UNICODE_SCRIPT_SYRIAC, 0x0711 }, |
| { G_UNICODE_SCRIPT_TAMIL, 0x0B82 }, |
| { G_UNICODE_SCRIPT_TELUGU, 0x0C03 }, |
| { G_UNICODE_SCRIPT_THAANA, 0x07B1 }, |
| { G_UNICODE_SCRIPT_THAI, 0x0E31 }, |
| { G_UNICODE_SCRIPT_TIBETAN, 0x0FD4 }, |
| { G_UNICODE_SCRIPT_CANADIAN_ABORIGINAL, 0x1400 }, |
| { G_UNICODE_SCRIPT_CANADIAN_ABORIGINAL, 0x1401 }, |
| { G_UNICODE_SCRIPT_YI, 0xA015 }, |
| { G_UNICODE_SCRIPT_TAGALOG, 0x1700 }, |
| { G_UNICODE_SCRIPT_HANUNOO, 0x1720 }, |
| { G_UNICODE_SCRIPT_BUHID, 0x1740 }, |
| { G_UNICODE_SCRIPT_TAGBANWA, 0x1760 }, |
| { G_UNICODE_SCRIPT_BRAILLE, 0x2800 }, |
| { G_UNICODE_SCRIPT_CYPRIOT, 0x10808 }, |
| { G_UNICODE_SCRIPT_LIMBU, 0x1932 }, |
| { G_UNICODE_SCRIPT_OSMANYA, 0x10480 }, |
| { G_UNICODE_SCRIPT_SHAVIAN, 0x10450 }, |
| { G_UNICODE_SCRIPT_LINEAR_B, 0x10000 }, |
| { G_UNICODE_SCRIPT_TAI_LE, 0x1950 }, |
| { G_UNICODE_SCRIPT_UGARITIC, 0x1039F }, |
| { G_UNICODE_SCRIPT_NEW_TAI_LUE, 0x1980 }, |
| { G_UNICODE_SCRIPT_BUGINESE, 0x1A1F }, |
| { G_UNICODE_SCRIPT_GLAGOLITIC, 0x2C00 }, |
| { G_UNICODE_SCRIPT_TIFINAGH, 0x2D6F }, |
| { G_UNICODE_SCRIPT_SYLOTI_NAGRI, 0xA800 }, |
| { G_UNICODE_SCRIPT_OLD_PERSIAN, 0x103D0 }, |
| { G_UNICODE_SCRIPT_KHAROSHTHI, 0x10A3F }, |
| { G_UNICODE_SCRIPT_UNKNOWN, 0x1111111 }, |
| { G_UNICODE_SCRIPT_BALINESE, 0x1B04 }, |
| { G_UNICODE_SCRIPT_CUNEIFORM, 0x12000 }, |
| { G_UNICODE_SCRIPT_PHOENICIAN, 0x10900 }, |
| { G_UNICODE_SCRIPT_PHAGS_PA, 0xA840 }, |
| { G_UNICODE_SCRIPT_NKO, 0x07C0 }, |
| { G_UNICODE_SCRIPT_KAYAH_LI, 0xA900 }, |
| { G_UNICODE_SCRIPT_LEPCHA, 0x1C00 }, |
| { G_UNICODE_SCRIPT_REJANG, 0xA930 }, |
| { G_UNICODE_SCRIPT_SUNDANESE, 0x1B80 }, |
| { G_UNICODE_SCRIPT_SAURASHTRA, 0xA880 }, |
| { G_UNICODE_SCRIPT_CHAM, 0xAA00 }, |
| { G_UNICODE_SCRIPT_OL_CHIKI, 0x1C50 }, |
| { G_UNICODE_SCRIPT_VAI, 0xA500 }, |
| { G_UNICODE_SCRIPT_CARIAN, 0x102A0 }, |
| { G_UNICODE_SCRIPT_LYCIAN, 0x10280 }, |
| { G_UNICODE_SCRIPT_LYDIAN, 0x1093F }, |
| { G_UNICODE_SCRIPT_AVESTAN, 0x10B00 }, |
| { G_UNICODE_SCRIPT_BAMUM, 0xA6A0 }, |
| { G_UNICODE_SCRIPT_EGYPTIAN_HIEROGLYPHS, 0x13000 }, |
| { G_UNICODE_SCRIPT_IMPERIAL_ARAMAIC, 0x10840 }, |
| { G_UNICODE_SCRIPT_INSCRIPTIONAL_PAHLAVI, 0x10B60 }, |
| { G_UNICODE_SCRIPT_INSCRIPTIONAL_PARTHIAN, 0x10B40 }, |
| { G_UNICODE_SCRIPT_JAVANESE, 0xA980 }, |
| { G_UNICODE_SCRIPT_KAITHI, 0x11082 }, |
| { G_UNICODE_SCRIPT_LISU, 0xA4D0 }, |
| { G_UNICODE_SCRIPT_MEETEI_MAYEK, 0xABE5 }, |
| { G_UNICODE_SCRIPT_OLD_SOUTH_ARABIAN, 0x10A60 }, |
| { G_UNICODE_SCRIPT_OLD_TURKIC, 0x10C00 }, |
| { G_UNICODE_SCRIPT_SAMARITAN, 0x0800 }, |
| { G_UNICODE_SCRIPT_TAI_THAM, 0x1A20 }, |
| { G_UNICODE_SCRIPT_TAI_VIET, 0xAA80 }, |
| { G_UNICODE_SCRIPT_BATAK, 0x1BC0 }, |
| { G_UNICODE_SCRIPT_BRAHMI, 0x11000 }, |
| { G_UNICODE_SCRIPT_MANDAIC, 0x0840 }, |
| { G_UNICODE_SCRIPT_CHAKMA, 0x11100 }, |
| { G_UNICODE_SCRIPT_MEROITIC_CURSIVE, 0x109A0 }, |
| { G_UNICODE_SCRIPT_MEROITIC_HIEROGLYPHS, 0x10980 }, |
| { G_UNICODE_SCRIPT_MIAO, 0x16F00 }, |
| { G_UNICODE_SCRIPT_SHARADA, 0x11180 }, |
| { G_UNICODE_SCRIPT_SORA_SOMPENG, 0x110D0 }, |
| { G_UNICODE_SCRIPT_TAKRI, 0x11680 }, |
| { G_UNICODE_SCRIPT_BASSA_VAH, 0x16AD0 }, |
| { G_UNICODE_SCRIPT_CAUCASIAN_ALBANIAN, 0x10530 }, |
| { G_UNICODE_SCRIPT_DUPLOYAN, 0x1BC00 }, |
| { G_UNICODE_SCRIPT_ELBASAN, 0x10500 }, |
| { G_UNICODE_SCRIPT_GRANTHA, 0x11301 }, |
| { G_UNICODE_SCRIPT_KHOJKI, 0x11200 }, |
| { G_UNICODE_SCRIPT_KHUDAWADI, 0x112B0 }, |
| { G_UNICODE_SCRIPT_LINEAR_A, 0x10600 }, |
| { G_UNICODE_SCRIPT_MAHAJANI, 0x11150 }, |
| { G_UNICODE_SCRIPT_MANICHAEAN, 0x10AC0 }, |
| { G_UNICODE_SCRIPT_MENDE_KIKAKUI, 0x1E800 }, |
| { G_UNICODE_SCRIPT_MODI, 0x11600 }, |
| { G_UNICODE_SCRIPT_MRO, 0x16A40 }, |
| { G_UNICODE_SCRIPT_NABATAEAN, 0x10880 }, |
| { G_UNICODE_SCRIPT_OLD_NORTH_ARABIAN, 0x10A80 }, |
| { G_UNICODE_SCRIPT_OLD_PERMIC, 0x10350 }, |
| { G_UNICODE_SCRIPT_PAHAWH_HMONG, 0x16B00 }, |
| { G_UNICODE_SCRIPT_PALMYRENE, 0x10860 }, |
| { G_UNICODE_SCRIPT_PAU_CIN_HAU, 0x11AC0 }, |
| { G_UNICODE_SCRIPT_PSALTER_PAHLAVI, 0x10B80 }, |
| { G_UNICODE_SCRIPT_SIDDHAM, 0x11580 }, |
| { G_UNICODE_SCRIPT_TIRHUTA, 0x11480 }, |
| { G_UNICODE_SCRIPT_WARANG_CITI, 0x118A0 }, |
| { G_UNICODE_SCRIPT_CHEROKEE, 0x0AB71 }, |
| { G_UNICODE_SCRIPT_HATRAN, 0x108E0 }, |
| { G_UNICODE_SCRIPT_OLD_HUNGARIAN, 0x10C80 }, |
| { G_UNICODE_SCRIPT_MULTANI, 0x11280 }, |
| { G_UNICODE_SCRIPT_AHOM, 0x11700 }, |
| { G_UNICODE_SCRIPT_CUNEIFORM, 0x12480 }, |
| { G_UNICODE_SCRIPT_ANATOLIAN_HIEROGLYPHS, 0x14400 }, |
| { G_UNICODE_SCRIPT_SIGNWRITING, 0x1D800 }, |
| { G_UNICODE_SCRIPT_ADLAM, 0x1E900 }, |
| { G_UNICODE_SCRIPT_BHAIKSUKI, 0x11C00 }, |
| { G_UNICODE_SCRIPT_MARCHEN, 0x11C70 }, |
| { G_UNICODE_SCRIPT_NEWA, 0x11400 }, |
| { G_UNICODE_SCRIPT_OSAGE, 0x104B0 }, |
| { G_UNICODE_SCRIPT_TANGUT, 0x16FE0 }, |
| { G_UNICODE_SCRIPT_MASARAM_GONDI, 0x11D00 }, |
| { G_UNICODE_SCRIPT_NUSHU, 0x1B170 }, |
| { G_UNICODE_SCRIPT_SOYOMBO, 0x11A50 }, |
| { G_UNICODE_SCRIPT_ZANABAZAR_SQUARE, 0x11A00 }, |
| }; |
| for (i = 0; i < G_N_ELEMENTS (examples); i++) |
| g_assert_cmpint (g_unichar_get_script (examples[i].c), ==, examples[i].script); |
| } |
| |
| static void |
| test_combining_class (void) |
| { |
| guint i; |
| struct { |
| gint class; |
| gunichar c; |
| } examples[] = { |
| { 0, 0x0020 }, |
| { 1, 0x0334 }, |
| { 7, 0x093C }, |
| { 8, 0x3099 }, |
| { 9, 0x094D }, |
| { 10, 0x05B0 }, |
| { 11, 0x05B1 }, |
| { 12, 0x05B2 }, |
| { 13, 0x05B3 }, |
| { 14, 0x05B4 }, |
| { 15, 0x05B5 }, |
| { 16, 0x05B6 }, |
| { 17, 0x05B7 }, |
| { 18, 0x05B8 }, |
| { 19, 0x05B9 }, |
| { 20, 0x05BB }, |
| { 21, 0x05BC }, |
| { 22, 0x05BD }, |
| { 23, 0x05BF }, |
| { 24, 0x05C1 }, |
| { 25, 0x05C2 }, |
| { 26, 0xFB1E }, |
| { 27, 0x064B }, |
| { 28, 0x064C }, |
| { 29, 0x064D }, |
| /* ... */ |
| { 228, 0x05AE }, |
| { 230, 0x0300 }, |
| { 232, 0x302C }, |
| { 233, 0x0362 }, |
| { 234, 0x0360 }, |
| { 234, 0x1DCD }, |
| { 240, 0x0345 } |
| }; |
| for (i = 0; i < G_N_ELEMENTS (examples); i++) |
| { |
| g_assert_cmpint (g_unichar_combining_class (examples[i].c), ==, examples[i].class); |
| } |
| } |
| |
| static void |
| test_mirror (void) |
| { |
| gunichar mirror; |
| |
| g_assert (g_unichar_get_mirror_char ('(', &mirror)); |
| g_assert_cmpint (mirror, ==, ')'); |
| g_assert (g_unichar_get_mirror_char (')', &mirror)); |
| g_assert_cmpint (mirror, ==, '('); |
| g_assert (g_unichar_get_mirror_char ('{', &mirror)); |
| g_assert_cmpint (mirror, ==, '}'); |
| g_assert (g_unichar_get_mirror_char ('}', &mirror)); |
| g_assert_cmpint (mirror, ==, '{'); |
| g_assert (g_unichar_get_mirror_char (0x208D, &mirror)); |
| g_assert_cmpint (mirror, ==, 0x208E); |
| g_assert (g_unichar_get_mirror_char (0x208E, &mirror)); |
| g_assert_cmpint (mirror, ==, 0x208D); |
| g_assert (!g_unichar_get_mirror_char ('a', &mirror)); |
| } |
| |
| static void |
| test_mark (void) |
| { |
| g_assert (g_unichar_ismark (0x0903)); |
| g_assert (g_unichar_ismark (0x20DD)); |
| g_assert (g_unichar_ismark (0xA806)); |
| g_assert (!g_unichar_ismark ('a')); |
| } |
| |
| static void |
| test_zerowidth (void) |
| { |
| g_assert (!g_unichar_iszerowidth (0x00AD)); |
| g_assert (!g_unichar_iszerowidth (0x00AD)); |
| g_assert (!g_unichar_iszerowidth (0x115F)); |
| g_assert (g_unichar_iszerowidth (0x1160)); |
| g_assert (g_unichar_iszerowidth (0x11AA)); |
| g_assert (g_unichar_iszerowidth (0x11FF)); |
| g_assert (!g_unichar_iszerowidth (0x1200)); |
| g_assert (g_unichar_iszerowidth (0x200B)); |
| g_assert (g_unichar_iszerowidth (0x591)); |
| } |
| |
| static void |
| test_title (void) |
| { |
| g_assert (g_unichar_istitle (0x01c5)); |
| g_assert (g_unichar_istitle (0x1f88)); |
| g_assert (g_unichar_istitle (0x1fcc)); |
| g_assert (!g_unichar_istitle ('a')); |
| g_assert (!g_unichar_istitle ('A')); |
| |
| g_assert_cmphex (g_unichar_totitle (0x01c6), ==, 0x01c5); |
| g_assert_cmphex (g_unichar_totitle (0x01c4), ==, 0x01c5); |
| g_assert_cmphex (g_unichar_totitle (0x01c5), ==, 0x01c5); |
| g_assert_cmphex (g_unichar_totitle (0x1f80), ==, 0x1f88); |
| g_assert_cmphex (g_unichar_totitle (0x1f88), ==, 0x1f88); |
| g_assert_cmphex (g_unichar_totitle ('a'), ==, 'A'); |
| g_assert_cmphex (g_unichar_totitle ('A'), ==, 'A'); |
| } |
| |
| static void |
| test_cases (void) |
| { |
| g_assert_cmphex (g_unichar_toupper ('a'), ==, 'A'); |
| g_assert_cmphex (g_unichar_toupper ('A'), ==, 'A'); |
| g_assert_cmphex (g_unichar_toupper (0x01C5), ==, 0x01C4); |
| g_assert_cmphex (g_unichar_toupper (0x01C6), ==, 0x01C4); |
| g_assert_cmphex (g_unichar_tolower ('A'), ==, 'a'); |
| g_assert_cmphex (g_unichar_tolower ('a'), ==, 'a'); |
| g_assert_cmphex (g_unichar_tolower (0x01C4), ==, 0x01C6); |
| g_assert_cmphex (g_unichar_tolower (0x01C5), ==, 0x01C6); |
| g_assert_cmphex (g_unichar_tolower (0x1F8A), ==, 0x1F82); |
| g_assert_cmphex (g_unichar_totitle (0x1F8A), ==, 0x1F8A); |
| g_assert_cmphex (g_unichar_toupper (0x1F8A), ==, 0x1F8A); |
| g_assert_cmphex (g_unichar_tolower (0x1FB2), ==, 0x1FB2); |
| g_assert_cmphex (g_unichar_toupper (0x1FB2), ==, 0x1FB2); |
| } |
| |
| static void |
| test_defined (void) |
| { |
| g_assert (g_unichar_isdefined (0x0903)); |
| g_assert (g_unichar_isdefined (0x20DD)); |
| g_assert (g_unichar_isdefined (0x20BA)); |
| g_assert (g_unichar_isdefined (0xA806)); |
| g_assert (g_unichar_isdefined ('a')); |
| g_assert (!g_unichar_isdefined (0x10C49)); |
| g_assert (!g_unichar_isdefined (0x169D)); |
| } |
| |
| static void |
| test_wide (void) |
| { |
| guint i; |
| struct { |
| gunichar c; |
| enum { |
| NOT_WIDE, |
| WIDE_CJK, |
| WIDE |
| } wide; |
| } examples[] = { |
| /* Neutral */ |
| { 0x0000, NOT_WIDE }, |
| { 0x0483, NOT_WIDE }, |
| { 0x0641, NOT_WIDE }, |
| { 0xFFFC, NOT_WIDE }, |
| { 0x10000, NOT_WIDE }, |
| { 0xE0001, NOT_WIDE }, |
| { 0x2FFFE, NOT_WIDE }, |
| { 0x3FFFE, NOT_WIDE }, |
| |
| /* Narrow */ |
| { 0x0020, NOT_WIDE }, |
| { 0x0041, NOT_WIDE }, |
| { 0x27E6, NOT_WIDE }, |
| |
| /* Halfwidth */ |
| { 0x20A9, NOT_WIDE }, |
| { 0xFF61, NOT_WIDE }, |
| { 0xFF69, NOT_WIDE }, |
| { 0xFFEE, NOT_WIDE }, |
| |
| /* Ambiguous */ |
| { 0x00A1, WIDE_CJK }, |
| { 0x00BE, WIDE_CJK }, |
| { 0x02DD, WIDE_CJK }, |
| { 0x2020, WIDE_CJK }, |
| { 0xFFFD, WIDE_CJK }, |
| { 0x00A1, WIDE_CJK }, |
| { 0x1F100, WIDE_CJK }, |
| { 0xE0100, WIDE_CJK }, |
| { 0x100000, WIDE_CJK }, |
| { 0x10FFFD, WIDE_CJK }, |
| |
| /* Fullwidth */ |
| { 0x3000, WIDE }, |
| { 0xFF60, WIDE }, |
| |
| /* Wide */ |
| { 0x2329, WIDE }, |
| { 0x3001, WIDE }, |
| { 0xFE69, WIDE }, |
| { 0x30000, WIDE }, |
| { 0x3FFFD, WIDE }, |
| |
| /* Default Wide blocks */ |
| { 0x4DBF, WIDE }, |
| { 0x9FFF, WIDE }, |
| { 0xFAFF, WIDE }, |
| { 0x2A6DF, WIDE }, |
| { 0x2B73F, WIDE }, |
| { 0x2B81F, WIDE }, |
| { 0x2FA1F, WIDE }, |
| |
| /* Uniode-5.2 character additions */ |
| /* Wide */ |
| { 0x115F, WIDE }, |
| |
| /* Uniode-6.0 character additions */ |
| /* Wide */ |
| { 0x2B740, WIDE }, |
| { 0x1B000, WIDE }, |
| |
| { 0x111111, NOT_WIDE } |
| }; |
| |
| for (i = 0; i < G_N_ELEMENTS (examples); i++) |
| { |
| g_assert_cmpint (g_unichar_iswide (examples[i].c), ==, (examples[i].wide == WIDE)); |
| g_assert_cmpint (g_unichar_iswide_cjk (examples[i].c), ==, (examples[i].wide != NOT_WIDE)); |
| } |
| }; |
| |
| static void |
| test_compose (void) |
| { |
| gunichar ch; |
| |
| /* Not composable */ |
| g_assert (!g_unichar_compose (0x0041, 0x0042, &ch) && ch == 0); |
| g_assert (!g_unichar_compose (0x0041, 0, &ch) && ch == 0); |
| g_assert (!g_unichar_compose (0x0066, 0x0069, &ch) && ch == 0); |
| |
| /* Tricky non-composable */ |
| g_assert (!g_unichar_compose (0x0308, 0x0301, &ch) && ch == 0); /* !0x0344 */ |
| g_assert (!g_unichar_compose (0x0F71, 0x0F72, &ch) && ch == 0); /* !0x0F73 */ |
| |
| /* Singletons should not compose */ |
| g_assert (!g_unichar_compose (0x212B, 0, &ch) && ch == 0); |
| g_assert (!g_unichar_compose (0x00C5, 0, &ch) && ch == 0); |
| g_assert (!g_unichar_compose (0x2126, 0, &ch) && ch == 0); |
| g_assert (!g_unichar_compose (0x03A9, 0, &ch) && ch == 0); |
| |
| /* Pairs */ |
| g_assert (g_unichar_compose (0x0041, 0x030A, &ch) && ch == 0x00C5); |
| g_assert (g_unichar_compose (0x006F, 0x0302, &ch) && ch == 0x00F4); |
| g_assert (g_unichar_compose (0x1E63, 0x0307, &ch) && ch == 0x1E69); |
| g_assert (g_unichar_compose (0x0073, 0x0323, &ch) && ch == 0x1E63); |
| g_assert (g_unichar_compose (0x0064, 0x0307, &ch) && ch == 0x1E0B); |
| g_assert (g_unichar_compose (0x0064, 0x0323, &ch) && ch == 0x1E0D); |
| |
| /* Hangul */ |
| g_assert (g_unichar_compose (0xD4CC, 0x11B6, &ch) && ch == 0xD4DB); |
| g_assert (g_unichar_compose (0x1111, 0x1171, &ch) && ch == 0xD4CC); |
| g_assert (g_unichar_compose (0xCE20, 0x11B8, &ch) && ch == 0xCE31); |
| g_assert (g_unichar_compose (0x110E, 0x1173, &ch) && ch == 0xCE20); |
| } |
| |
| static void |
| test_decompose (void) |
| { |
| gunichar a, b; |
| |
| /* Not decomposable */ |
| g_assert (!g_unichar_decompose (0x0041, &a, &b) && a == 0x0041 && b == 0); |
| g_assert (!g_unichar_decompose (0xFB01, &a, &b) && a == 0xFB01 && b == 0); |
| |
| /* Singletons */ |
| g_assert (g_unichar_decompose (0x212B, &a, &b) && a == 0x00C5 && b == 0); |
| g_assert (g_unichar_decompose (0x2126, &a, &b) && a == 0x03A9 && b == 0); |
| |
| /* Tricky pairs */ |
| g_assert (g_unichar_decompose (0x0344, &a, &b) && a == 0x0308 && b == 0x0301); |
| g_assert (g_unichar_decompose (0x0F73, &a, &b) && a == 0x0F71 && b == 0x0F72); |
| |
| /* Pairs */ |
| g_assert (g_unichar_decompose (0x00C5, &a, &b) && a == 0x0041 && b == 0x030A); |
| g_assert (g_unichar_decompose (0x00F4, &a, &b) && a == 0x006F && b == 0x0302); |
| g_assert (g_unichar_decompose (0x1E69, &a, &b) && a == 0x1E63 && b == 0x0307); |
| g_assert (g_unichar_decompose (0x1E63, &a, &b) && a == 0x0073 && b == 0x0323); |
| g_assert (g_unichar_decompose (0x1E0B, &a, &b) && a == 0x0064 && b == 0x0307); |
| g_assert (g_unichar_decompose (0x1E0D, &a, &b) && a == 0x0064 && b == 0x0323); |
| |
| /* Hangul */ |
| g_assert (g_unichar_decompose (0xD4DB, &a, &b) && a == 0xD4CC && b == 0x11B6); |
| g_assert (g_unichar_decompose (0xD4CC, &a, &b) && a == 0x1111 && b == 0x1171); |
| g_assert (g_unichar_decompose (0xCE31, &a, &b) && a == 0xCE20 && b == 0x11B8); |
| g_assert (g_unichar_decompose (0xCE20, &a, &b) && a == 0x110E && b == 0x1173); |
| } |
| |
| static void |
| test_fully_decompose_canonical (void) |
| { |
| gunichar decomp[5]; |
| gsize len; |
| |
| #define TEST_DECOMP(ch, expected_len, a, b, c, d) \ |
| len = g_unichar_fully_decompose (ch, FALSE, decomp, G_N_ELEMENTS (decomp)); \ |
| g_assert_cmpint (expected_len, ==, len); \ |
| if (expected_len >= 1) g_assert_cmphex (decomp[0], ==, a); \ |
| if (expected_len >= 2) g_assert_cmphex (decomp[1], ==, b); \ |
| if (expected_len >= 3) g_assert_cmphex (decomp[2], ==, c); \ |
| if (expected_len >= 4) g_assert_cmphex (decomp[3], ==, d); \ |
| |
| #define TEST0(ch) TEST_DECOMP (ch, 1, ch, 0, 0, 0) |
| #define TEST1(ch, a) TEST_DECOMP (ch, 1, a, 0, 0, 0) |
| #define TEST2(ch, a, b) TEST_DECOMP (ch, 2, a, b, 0, 0) |
| #define TEST3(ch, a, b, c) TEST_DECOMP (ch, 3, a, b, c, 0) |
| #define TEST4(ch, a, b, c, d) TEST_DECOMP (ch, 4, a, b, c, d) |
| |
| /* Not decomposable */ |
| TEST0 (0x0041); |
| TEST0 (0xFB01); |
| |
| /* Singletons */ |
| TEST2 (0x212B, 0x0041, 0x030A); |
| TEST1 (0x2126, 0x03A9); |
| |
| /* Tricky pairs */ |
| TEST2 (0x0344, 0x0308, 0x0301); |
| TEST2 (0x0F73, 0x0F71, 0x0F72); |
| |
| /* General */ |
| TEST2 (0x00C5, 0x0041, 0x030A); |
| TEST2 (0x00F4, 0x006F, 0x0302); |
| TEST3 (0x1E69, 0x0073, 0x0323, 0x0307); |
| TEST2 (0x1E63, 0x0073, 0x0323); |
| TEST2 (0x1E0B, 0x0064, 0x0307); |
| TEST2 (0x1E0D, 0x0064, 0x0323); |
| |
| /* Hangul */ |
| TEST3 (0xD4DB, 0x1111, 0x1171, 0x11B6); |
| TEST2 (0xD4CC, 0x1111, 0x1171); |
| TEST3 (0xCE31, 0x110E, 0x1173, 0x11B8); |
| TEST2 (0xCE20, 0x110E, 0x1173); |
| |
| #undef TEST_DECOMP |
| } |
| |
| static void |
| test_canonical_decomposition (void) |
| { |
| gunichar *decomp; |
| gsize len; |
| |
| #define TEST_DECOMP(ch, expected_len, a, b, c, d) \ |
| decomp = g_unicode_canonical_decomposition (ch, &len); \ |
| g_assert_cmpint (expected_len, ==, len); \ |
| if (expected_len >= 1) g_assert_cmphex (decomp[0], ==, a); \ |
| if (expected_len >= 2) g_assert_cmphex (decomp[1], ==, b); \ |
| if (expected_len >= 3) g_assert_cmphex (decomp[2], ==, c); \ |
| if (expected_len >= 4) g_assert_cmphex (decomp[3], ==, d); \ |
| g_free (decomp); |
| |
| #define TEST0(ch) TEST_DECOMP (ch, 1, ch, 0, 0, 0) |
| #define TEST1(ch, a) TEST_DECOMP (ch, 1, a, 0, 0, 0) |
| #define TEST2(ch, a, b) TEST_DECOMP (ch, 2, a, b, 0, 0) |
| #define TEST3(ch, a, b, c) TEST_DECOMP (ch, 3, a, b, c, 0) |
| #define TEST4(ch, a, b, c, d) TEST_DECOMP (ch, 4, a, b, c, d) |
| |
| /* Not decomposable */ |
| TEST0 (0x0041); |
| TEST0 (0xFB01); |
| |
| /* Singletons */ |
| TEST2 (0x212B, 0x0041, 0x030A); |
| TEST1 (0x2126, 0x03A9); |
| |
| /* Tricky pairs */ |
| TEST2 (0x0344, 0x0308, 0x0301); |
| TEST2 (0x0F73, 0x0F71, 0x0F72); |
| |
| /* General */ |
| TEST2 (0x00C5, 0x0041, 0x030A); |
| TEST2 (0x00F4, 0x006F, 0x0302); |
| TEST3 (0x1E69, 0x0073, 0x0323, 0x0307); |
| TEST2 (0x1E63, 0x0073, 0x0323); |
| TEST2 (0x1E0B, 0x0064, 0x0307); |
| TEST2 (0x1E0D, 0x0064, 0x0323); |
| |
| /* Hangul */ |
| TEST3 (0xD4DB, 0x1111, 0x1171, 0x11B6); |
| TEST2 (0xD4CC, 0x1111, 0x1171); |
| TEST3 (0xCE31, 0x110E, 0x1173, 0x11B8); |
| TEST2 (0xCE20, 0x110E, 0x1173); |
| |
| #undef TEST_DECOMP |
| } |
| |
| static void |
| test_decompose_tail (void) |
| { |
| gunichar ch, a, b, c, d; |
| |
| /* Test that whenever a char ch decomposes into a and b, b itself |
| * won't decompose any further. */ |
| |
| for (ch = 0; ch < 0x110000; ch++) |
| if (g_unichar_decompose (ch, &a, &b)) |
| g_assert (!g_unichar_decompose (b, &c, &d)); |
| else |
| { |
| g_assert_cmpuint (a, ==, ch); |
| g_assert_cmpuint (b, ==, 0); |
| } |
| } |
| |
| static void |
| test_fully_decompose_len (void) |
| { |
| gunichar ch; |
| |
| /* Test that all canonical decompositions are at most 4 in length, |
| * and compatibility decompositions are at most 18 in length. |
| */ |
| |
| for (ch = 0; ch < 0x110000; ch++) { |
| g_assert_cmpint (g_unichar_fully_decompose (ch, FALSE, NULL, 0), <=, 4); |
| g_assert_cmpint (g_unichar_fully_decompose (ch, TRUE, NULL, 0), <=, 18); |
| } |
| } |
| |
| static void |
| test_iso15924 (void) |
| { |
| const struct { |
| GUnicodeScript script; |
| char four_letter_code[5]; |
| } data[] = { |
| { G_UNICODE_SCRIPT_COMMON, "Zyyy" }, |
| { G_UNICODE_SCRIPT_INHERITED, "Zinh" }, |
| { G_UNICODE_SCRIPT_ARABIC, "Arab" }, |
| { G_UNICODE_SCRIPT_ARMENIAN, "Armn" }, |
| { G_UNICODE_SCRIPT_BENGALI, "Beng" }, |
| { G_UNICODE_SCRIPT_BOPOMOFO, "Bopo" }, |
| { G_UNICODE_SCRIPT_CHEROKEE, "Cher" }, |
| { G_UNICODE_SCRIPT_COPTIC, "Copt" }, |
| { G_UNICODE_SCRIPT_CYRILLIC, "Cyrl" }, |
| { G_UNICODE_SCRIPT_DESERET, "Dsrt" }, |
| { G_UNICODE_SCRIPT_DEVANAGARI, "Deva" }, |
| { G_UNICODE_SCRIPT_ETHIOPIC, "Ethi" }, |
| { G_UNICODE_SCRIPT_GEORGIAN, "Geor" }, |
| { G_UNICODE_SCRIPT_GOTHIC, "Goth" }, |
| { G_UNICODE_SCRIPT_GREEK, "Grek" }, |
| { G_UNICODE_SCRIPT_GUJARATI, "Gujr" }, |
| { G_UNICODE_SCRIPT_GURMUKHI, "Guru" }, |
| { G_UNICODE_SCRIPT_HAN, "Hani" }, |
| { G_UNICODE_SCRIPT_HANGUL, "Hang" }, |
| { G_UNICODE_SCRIPT_HEBREW, "Hebr" }, |
| { G_UNICODE_SCRIPT_HIRAGANA, "Hira" }, |
| { G_UNICODE_SCRIPT_KANNADA, "Knda" }, |
| { G_UNICODE_SCRIPT_KATAKANA, "Kana" }, |
| { G_UNICODE_SCRIPT_KHMER, "Khmr" }, |
| { G_UNICODE_SCRIPT_LAO, "Laoo" }, |
| { G_UNICODE_SCRIPT_LATIN, "Latn" }, |
| { G_UNICODE_SCRIPT_MALAYALAM, "Mlym" }, |
| { G_UNICODE_SCRIPT_MONGOLIAN, "Mong" }, |
| { G_UNICODE_SCRIPT_MYANMAR, "Mymr" }, |
| { G_UNICODE_SCRIPT_OGHAM, "Ogam" }, |
| { G_UNICODE_SCRIPT_OLD_ITALIC, "Ital" }, |
| { G_UNICODE_SCRIPT_ORIYA, "Orya" }, |
| { G_UNICODE_SCRIPT_RUNIC, "Runr" }, |
| { G_UNICODE_SCRIPT_SINHALA, "Sinh" }, |
| { G_UNICODE_SCRIPT_SYRIAC, "Syrc" }, |
| { G_UNICODE_SCRIPT_TAMIL, "Taml" }, |
| { G_UNICODE_SCRIPT_TELUGU, "Telu" }, |
| { G_UNICODE_SCRIPT_THAANA, "Thaa" }, |
| { G_UNICODE_SCRIPT_THAI, "Thai" }, |
| { G_UNICODE_SCRIPT_TIBETAN, "Tibt" }, |
| { G_UNICODE_SCRIPT_CANADIAN_ABORIGINAL, "Cans" }, |
| { G_UNICODE_SCRIPT_YI, "Yiii" }, |
| { G_UNICODE_SCRIPT_TAGALOG, "Tglg" }, |
| { G_UNICODE_SCRIPT_HANUNOO, "Hano" }, |
| { G_UNICODE_SCRIPT_BUHID, "Buhd" }, |
| { G_UNICODE_SCRIPT_TAGBANWA, "Tagb" }, |
| |
| /* Unicode-4.0 additions */ |
| { G_UNICODE_SCRIPT_BRAILLE, "Brai" }, |
| { G_UNICODE_SCRIPT_CYPRIOT, "Cprt" }, |
| { G_UNICODE_SCRIPT_LIMBU, "Limb" }, |
| { G_UNICODE_SCRIPT_OSMANYA, "Osma" }, |
| { G_UNICODE_SCRIPT_SHAVIAN, "Shaw" }, |
| { G_UNICODE_SCRIPT_LINEAR_B, "Linb" }, |
| { G_UNICODE_SCRIPT_TAI_LE, "Tale" }, |
| { G_UNICODE_SCRIPT_UGARITIC, "Ugar" }, |
| |
| /* Unicode-4.1 additions */ |
| { G_UNICODE_SCRIPT_NEW_TAI_LUE, "Talu" }, |
| { G_UNICODE_SCRIPT_BUGINESE, "Bugi" }, |
| { G_UNICODE_SCRIPT_GLAGOLITIC, "Glag" }, |
| { G_UNICODE_SCRIPT_TIFINAGH, "Tfng" }, |
| { G_UNICODE_SCRIPT_SYLOTI_NAGRI, "Sylo" }, |
| { G_UNICODE_SCRIPT_OLD_PERSIAN, "Xpeo" }, |
| { G_UNICODE_SCRIPT_KHAROSHTHI, "Khar" }, |
| |
| /* Unicode-5.0 additions */ |
| { G_UNICODE_SCRIPT_UNKNOWN, "Zzzz" }, |
| { G_UNICODE_SCRIPT_BALINESE, "Bali" }, |
| { G_UNICODE_SCRIPT_CUNEIFORM, "Xsux" }, |
| { G_UNICODE_SCRIPT_PHOENICIAN, "Phnx" }, |
| { G_UNICODE_SCRIPT_PHAGS_PA, "Phag" }, |
| { G_UNICODE_SCRIPT_NKO, "Nkoo" }, |
| |
| /* Unicode-5.1 additions */ |
| { G_UNICODE_SCRIPT_KAYAH_LI, "Kali" }, |
| { G_UNICODE_SCRIPT_LEPCHA, "Lepc" }, |
| { G_UNICODE_SCRIPT_REJANG, "Rjng" }, |
| { G_UNICODE_SCRIPT_SUNDANESE, "Sund" }, |
| { G_UNICODE_SCRIPT_SAURASHTRA, "Saur" }, |
| { G_UNICODE_SCRIPT_CHAM, "Cham" }, |
| { G_UNICODE_SCRIPT_OL_CHIKI, "Olck" }, |
| { G_UNICODE_SCRIPT_VAI, "Vaii" }, |
| { G_UNICODE_SCRIPT_CARIAN, "Cari" }, |
| { G_UNICODE_SCRIPT_LYCIAN, "Lyci" }, |
| { G_UNICODE_SCRIPT_LYDIAN, "Lydi" }, |
| |
| /* Unicode-5.2 additions */ |
| { G_UNICODE_SCRIPT_AVESTAN, "Avst" }, |
| { G_UNICODE_SCRIPT_BAMUM, "Bamu" }, |
| { G_UNICODE_SCRIPT_EGYPTIAN_HIEROGLYPHS, "Egyp" }, |
| { G_UNICODE_SCRIPT_IMPERIAL_ARAMAIC, "Armi" }, |
| { G_UNICODE_SCRIPT_INSCRIPTIONAL_PAHLAVI, "Phli" }, |
| { G_UNICODE_SCRIPT_INSCRIPTIONAL_PARTHIAN, "Prti" }, |
| { G_UNICODE_SCRIPT_JAVANESE, "Java" }, |
| { G_UNICODE_SCRIPT_KAITHI, "Kthi" }, |
| { G_UNICODE_SCRIPT_LISU, "Lisu" }, |
| { G_UNICODE_SCRIPT_MEETEI_MAYEK, "Mtei" }, |
| { G_UNICODE_SCRIPT_OLD_SOUTH_ARABIAN, "Sarb" }, |
| { G_UNICODE_SCRIPT_OLD_TURKIC, "Orkh" }, |
| { G_UNICODE_SCRIPT_SAMARITAN, "Samr" }, |
| { G_UNICODE_SCRIPT_TAI_THAM, "Lana" }, |
| { G_UNICODE_SCRIPT_TAI_VIET, "Tavt" }, |
| |
| /* Unicode-6.0 additions */ |
| { G_UNICODE_SCRIPT_BATAK, "Batk" }, |
| { G_UNICODE_SCRIPT_BRAHMI, "Brah" }, |
| { G_UNICODE_SCRIPT_MANDAIC, "Mand" }, |
| |
| /* Unicode-6.1 additions */ |
| { G_UNICODE_SCRIPT_CHAKMA, "Cakm" }, |
| { G_UNICODE_SCRIPT_MEROITIC_CURSIVE, "Merc" }, |
| { G_UNICODE_SCRIPT_MEROITIC_HIEROGLYPHS, "Mero" }, |
| { G_UNICODE_SCRIPT_MIAO, "Plrd" }, |
| { G_UNICODE_SCRIPT_SHARADA, "Shrd" }, |
| { G_UNICODE_SCRIPT_SORA_SOMPENG, "Sora" }, |
| { G_UNICODE_SCRIPT_TAKRI, "Takr" }, |
| |
| /* Unicode 7.0 additions */ |
| { G_UNICODE_SCRIPT_BASSA_VAH, "Bass" }, |
| { G_UNICODE_SCRIPT_CAUCASIAN_ALBANIAN, "Aghb" }, |
| { G_UNICODE_SCRIPT_DUPLOYAN, "Dupl" }, |
| { G_UNICODE_SCRIPT_ELBASAN, "Elba" }, |
| { G_UNICODE_SCRIPT_GRANTHA, "Gran" }, |
| { G_UNICODE_SCRIPT_KHOJKI, "Khoj" }, |
| { G_UNICODE_SCRIPT_KHUDAWADI, "Sind" }, |
| { G_UNICODE_SCRIPT_LINEAR_A, "Lina" }, |
| { G_UNICODE_SCRIPT_MAHAJANI, "Mahj" }, |
| { G_UNICODE_SCRIPT_MANICHAEAN, "Manu" }, |
| { G_UNICODE_SCRIPT_MENDE_KIKAKUI, "Mend" }, |
| { G_UNICODE_SCRIPT_MODI, "Modi" }, |
| { G_UNICODE_SCRIPT_MRO, "Mroo" }, |
| { G_UNICODE_SCRIPT_NABATAEAN, "Nbat" }, |
| { G_UNICODE_SCRIPT_OLD_NORTH_ARABIAN, "Narb" }, |
| { G_UNICODE_SCRIPT_OLD_PERMIC, "Perm" }, |
| { G_UNICODE_SCRIPT_PAHAWH_HMONG, "Hmng" }, |
| { G_UNICODE_SCRIPT_PALMYRENE, "Palm" }, |
| { G_UNICODE_SCRIPT_PAU_CIN_HAU, "Pauc" }, |
| { G_UNICODE_SCRIPT_PSALTER_PAHLAVI, "Phlp" }, |
| { G_UNICODE_SCRIPT_SIDDHAM, "Sidd" }, |
| { G_UNICODE_SCRIPT_TIRHUTA, "Tirh" }, |
| { G_UNICODE_SCRIPT_WARANG_CITI, "Wara" }, |
| |
| /* Unicode 8.0 additions */ |
| { G_UNICODE_SCRIPT_AHOM, "Ahom" }, |
| { G_UNICODE_SCRIPT_ANATOLIAN_HIEROGLYPHS, "Hluw" }, |
| { G_UNICODE_SCRIPT_HATRAN, "Hatr" }, |
| { G_UNICODE_SCRIPT_MULTANI, "Mult" }, |
| { G_UNICODE_SCRIPT_OLD_HUNGARIAN, "Hung" }, |
| { G_UNICODE_SCRIPT_SIGNWRITING, "Sgnw" }, |
| |
| /* Unicode 9.0 additions */ |
| { G_UNICODE_SCRIPT_ADLAM, "Adlm" }, |
| { G_UNICODE_SCRIPT_BHAIKSUKI, "Bhks" }, |
| { G_UNICODE_SCRIPT_MARCHEN, "Marc" }, |
| { G_UNICODE_SCRIPT_NEWA, "Newa" }, |
| { G_UNICODE_SCRIPT_OSAGE, "Osge" }, |
| { G_UNICODE_SCRIPT_TANGUT, "Tang" }, |
| |
| /* Unicode 10.0 additions */ |
| { G_UNICODE_SCRIPT_MASARAM_GONDI, "Gonm" }, |
| { G_UNICODE_SCRIPT_NUSHU, "Nshu" }, |
| { G_UNICODE_SCRIPT_SOYOMBO, "Soyo" }, |
| { G_UNICODE_SCRIPT_ZANABAZAR_SQUARE, "Zanb" }, |
| }; |
| guint i; |
| |
| g_assert_cmphex (0, ==, g_unicode_script_to_iso15924 (G_UNICODE_SCRIPT_INVALID_CODE)); |
| g_assert_cmphex (0x5A7A7A7A, ==, g_unicode_script_to_iso15924 (1000)); |
| g_assert_cmphex (0x41726162, ==, g_unicode_script_to_iso15924 (G_UNICODE_SCRIPT_ARABIC)); |
| |
| g_assert_cmphex (G_UNICODE_SCRIPT_INVALID_CODE, ==, g_unicode_script_from_iso15924 (0)); |
| g_assert_cmphex (G_UNICODE_SCRIPT_UNKNOWN, ==, g_unicode_script_from_iso15924 (0x12345678)); |
| |
| #define PACK(a,b,c,d) ((guint32)((((guint8)(a))<<24)|(((guint8)(b))<<16)|(((guint8)(c))<<8)|((guint8)(d)))) |
| |
| for (i = 0; i < G_N_ELEMENTS (data); i++) |
| { |
| guint32 code = PACK (data[i].four_letter_code[0], |
| data[i].four_letter_code[1], |
| data[i].four_letter_code[2], |
| data[i].four_letter_code[3]); |
| |
| g_assert_cmphex (g_unicode_script_to_iso15924 (data[i].script), ==, code); |
| g_assert_cmpint (g_unicode_script_from_iso15924 (code), ==, data[i].script); |
| } |
| |
| #undef PACK |
| } |
| |
| int |
| main (int argc, |
| char *argv[]) |
| { |
| g_test_init (&argc, &argv, NULL); |
| |
| g_test_add_func ("/unicode/validate", test_unichar_validate); |
| g_test_add_func ("/unicode/character-type", test_unichar_character_type); |
| g_test_add_func ("/unicode/break-type", test_unichar_break_type); |
| g_test_add_func ("/unicode/script", test_unichar_script); |
| g_test_add_func ("/unicode/combining-class", test_combining_class); |
| g_test_add_func ("/unicode/mirror", test_mirror); |
| g_test_add_func ("/unicode/mark", test_mark); |
| g_test_add_func ("/unicode/title", test_title); |
| g_test_add_func ("/unicode/zero-width", test_zerowidth); |
| g_test_add_func ("/unicode/defined", test_defined); |
| g_test_add_func ("/unicode/wide", test_wide); |
| g_test_add_func ("/unicode/compose", test_compose); |
| g_test_add_func ("/unicode/decompose", test_decompose); |
| g_test_add_func ("/unicode/fully-decompose-canonical", test_fully_decompose_canonical); |
| g_test_add_func ("/unicode/canonical-decomposition", test_canonical_decomposition); |
| g_test_add_func ("/unicode/decompose-tail", test_decompose_tail); |
| g_test_add_func ("/unicode/fully-decompose-len", test_fully_decompose_len); |
| g_test_add_func ("/unicode/iso15924", test_iso15924); |
| g_test_add_func ("/unicode/cases", test_cases); |
| |
| return g_test_run(); |
| } |