summaryrefslogtreecommitdiff
path: root/tests/uninorm/test-u8-nfd.c
diff options
context:
space:
mode:
authorAndreas Rottmann <a.rottmann@gmx.at>2009-09-14 12:32:44 +0200
committerAndreas Rottmann <a.rottmann@gmx.at>2009-09-14 12:32:44 +0200
commitfa095a4504cbe668e4244547e2c141597bea4ecf (patch)
tree06135820a286ffec47804e75fbf8a147e92acd2e /tests/uninorm/test-u8-nfd.c
Imported Upstream version 0.9.1upstream/0.9.1
Diffstat (limited to 'tests/uninorm/test-u8-nfd.c')
-rw-r--r--tests/uninorm/test-u8-nfd.c358
1 files changed, 358 insertions, 0 deletions
diff --git a/tests/uninorm/test-u8-nfd.c b/tests/uninorm/test-u8-nfd.c
new file mode 100644
index 00000000..d1d840ac
--- /dev/null
+++ b/tests/uninorm/test-u8-nfd.c
@@ -0,0 +1,358 @@
+/* Test of canonical decomposition of UTF-8 strings.
+ Copyright (C) 2009 Free Software Foundation, Inc.
+
+ This program is free software: you can redistribute it and/or modify
+ it under the terms of the GNU General Public License as published by
+ the Free Software Foundation; either version 3 of the License, or
+ (at your option) any later version.
+
+ This program 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. See the
+ GNU General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with this program. If not, see <http://www.gnu.org/licenses/>. */
+
+/* Written by Bruno Haible <bruno@clisp.org>, 2009. */
+
+#include <config.h>
+
+#if GNULIB_UNINORM_U8_NORMALIZE
+
+#include "uninorm.h"
+
+#include <signal.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+
+#include "unistr.h"
+
+#define SIZEOF(array) (sizeof (array) / sizeof (array[0]))
+#define ASSERT(expr) \
+ do \
+ { \
+ if (!(expr)) \
+ { \
+ fprintf (stderr, "%s:%d: assertion failed\n", __FILE__, __LINE__); \
+ fflush (stderr); \
+ abort (); \
+ } \
+ } \
+ while (0)
+
+static int
+check (const uint8_t *input, size_t input_length,
+ const uint8_t *expected, size_t expected_length)
+{
+ size_t length;
+ uint8_t *result;
+
+ /* Test return conventions with resultbuf == NULL. */
+ result = u8_normalize (UNINORM_NFD, input, input_length, NULL, &length);
+ if (!(result != NULL))
+ return 1;
+ if (!(length == expected_length))
+ return 2;
+ if (!(u8_cmp (result, expected, expected_length) == 0))
+ return 3;
+ free (result);
+
+ /* Test return conventions with resultbuf too small. */
+ if (expected_length > 0)
+ {
+ uint8_t *preallocated;
+
+ length = expected_length - 1;
+ preallocated = (uint8_t *) malloc (length * sizeof (uint8_t));
+ result = u8_normalize (UNINORM_NFD, input, input_length, preallocated, &length);
+ if (!(result != NULL))
+ return 4;
+ if (!(result != preallocated))
+ return 5;
+ if (!(length == expected_length))
+ return 6;
+ if (!(u8_cmp (result, expected, expected_length) == 0))
+ return 7;
+ free (result);
+ free (preallocated);
+ }
+
+ /* Test return conventions with resultbuf large enough. */
+ {
+ uint8_t *preallocated;
+
+ length = expected_length;
+ preallocated = (uint8_t *) malloc (length * sizeof (uint8_t));
+ result = u8_normalize (UNINORM_NFD, input, input_length, preallocated, &length);
+ if (!(result != NULL))
+ return 8;
+ if (!(preallocated == NULL || result == preallocated))
+ return 9;
+ if (!(length == expected_length))
+ return 10;
+ if (!(u8_cmp (result, expected, expected_length) == 0))
+ return 11;
+ free (preallocated);
+ }
+
+ return 0;
+}
+
+void
+test_u8_nfd (void)
+{
+ { /* Empty string. */
+ ASSERT (check (NULL, 0, NULL, 0) == 0);
+ }
+ { /* SPACE */
+ static const uint8_t input[] = { 0x20 };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* LATIN CAPITAL LETTER A WITH DIAERESIS */
+ static const uint8_t input[] = { 0xC3, 0x84 };
+ static const uint8_t expected[] = { 0x41, 0xCC, 0x88 };
+ ASSERT (check (input, SIZEOF (input), expected, SIZEOF (expected)) == 0);
+ }
+
+ { /* LATIN CAPITAL LETTER A WITH DIAERESIS AND MACRON */
+ static const uint8_t input[] = { 0xC7, 0x9E };
+ static const uint8_t expected[] = { 0x41, 0xCC, 0x88, 0xCC, 0x84 };
+ ASSERT (check (input, SIZEOF (input), expected, SIZEOF (expected)) == 0);
+ }
+
+ { /* GREEK DIALYTIKA AND PERISPOMENI */
+ static const uint8_t input[] = { 0xE1, 0xBF, 0x81 };
+ static const uint8_t expected[] = { 0xC2, 0xA8, 0xCD, 0x82 };
+ ASSERT (check (input, SIZEOF (input), expected, SIZEOF (expected)) == 0);
+ }
+
+ { /* SCRIPT SMALL L */
+ static const uint8_t input[] = { 0xE2, 0x84, 0x93 };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* NO-BREAK SPACE */
+ static const uint8_t input[] = { 0xC2, 0xA0 };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* ARABIC LETTER VEH INITIAL FORM */
+ static const uint8_t input[] = { 0xEF, 0xAD, 0xAC };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* ARABIC LETTER VEH MEDIAL FORM */
+ static const uint8_t input[] = { 0xEF, 0xAD, 0xAD };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* ARABIC LETTER VEH FINAL FORM */
+ static const uint8_t input[] = { 0xEF, 0xAD, 0xAB };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* ARABIC LETTER VEH ISOLATED FORM */
+ static const uint8_t input[] = { 0xEF, 0xAD, 0xAA };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* CIRCLED NUMBER FIFTEEN */
+ static const uint8_t input[] = { 0xE2, 0x91, 0xAE };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* TRADE MARK SIGN */
+ static const uint8_t input[] = { 0xE2, 0x84, 0xA2 };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* LATIN SUBSCRIPT SMALL LETTER I */
+ static const uint8_t input[] = { 0xE1, 0xB5, 0xA2 };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* PRESENTATION FORM FOR VERTICAL LEFT PARENTHESIS */
+ static const uint8_t input[] = { 0xEF, 0xB8, 0xB5 };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* FULLWIDTH LATIN CAPITAL LETTER A */
+ static const uint8_t input[] = { 0xEF, 0xBC, 0xA1 };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* HALFWIDTH IDEOGRAPHIC COMMA */
+ static const uint8_t input[] = { 0xEF, 0xBD, 0xA4 };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* SMALL IDEOGRAPHIC COMMA */
+ static const uint8_t input[] = { 0xEF, 0xB9, 0x91 };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* SQUARE MHZ */
+ static const uint8_t input[] = { 0xE3, 0x8E, 0x92 };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* VULGAR FRACTION THREE EIGHTHS */
+ static const uint8_t input[] = { 0xE2, 0x85, 0x9C };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* MICRO SIGN */
+ static const uint8_t input[] = { 0xC2, 0xB5 };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* ARABIC LIGATURE SALLALLAHOU ALAYHE WASALLAM */
+ static const uint8_t input[] = { 0xEF, 0xB7, 0xBA };
+ ASSERT (check (input, SIZEOF (input), input, SIZEOF (input)) == 0);
+ }
+
+ { /* HANGUL SYLLABLE GEUL */
+ static const uint8_t input[] = { 0xEA, 0xB8, 0x80 };
+ static const uint8_t expected[] =
+ { 0xE1, 0x84, 0x80, 0xE1, 0x85, 0xB3, 0xE1, 0x86, 0xAF };
+ ASSERT (check (input, SIZEOF (input), expected, SIZEOF (expected)) == 0);
+ }
+
+ { /* HANGUL SYLLABLE GEU */
+ static const uint8_t input[] = { 0xEA, 0xB7, 0xB8 };
+ static const uint8_t expected[] = { 0xE1, 0x84, 0x80, 0xE1, 0x85, 0xB3 };
+ ASSERT (check (input, SIZEOF (input), expected, SIZEOF (expected)) == 0);
+ }
+
+ { /* "Grüß Gott. Здравствуйте! x=(-b±sqrt(b²-4ac))/(2a) 日本語,中文,한글" */
+ static const uint8_t input[] =
+ { 'G', 'r', 0xC3, 0xBC, 0xC3, 0x9F, ' ', 'G', 'o', 't', 't', '.',
+ ' ', 0xD0, 0x97, 0xD0, 0xB4, 0xD1, 0x80, 0xD0, 0xB0, 0xD0, 0xB2, 0xD1,
+ 0x81, 0xD1, 0x82, 0xD0, 0xB2, 0xD1, 0x83, 0xD0, 0xB9,
+ 0xD1, 0x82, 0xD0, 0xB5, '!', ' ', 'x', '=', '(', '-', 'b', 0xC2, 0xB1,
+ 's', 'q', 'r', 't', '(', 'b', 0xC2, 0xB2, '-', '4', 'a', 'c', ')', ')',
+ '/', '(', '2', 'a', ')', ' ', ' ', 0xE6, 0x97, 0xA5, 0xE6, 0x9C, 0xAC,
+ 0xE8, 0xAA, 0x9E, ',', 0xE4, 0xB8, 0xAD, 0xE6, 0x96, 0x87, ',',
+ 0xED, 0x95, 0x9C,
+ 0xEA, 0xB8, 0x80, '\n'
+ };
+ static const uint8_t expected[] =
+ { 'G', 'r', 0x75, 0xCC, 0x88, 0xC3, 0x9F, ' ', 'G', 'o', 't', 't', '.',
+ ' ', 0xD0, 0x97, 0xD0, 0xB4, 0xD1, 0x80, 0xD0, 0xB0, 0xD0, 0xB2, 0xD1,
+ 0x81, 0xD1, 0x82, 0xD0, 0xB2, 0xD1, 0x83, 0xD0, 0xB8, 0xCC, 0x86,
+ 0xD1, 0x82, 0xD0, 0xB5, '!', ' ', 'x', '=', '(', '-', 'b', 0xC2, 0xB1,
+ 's', 'q', 'r', 't', '(', 'b', 0xC2, 0xB2, '-', '4', 'a', 'c', ')', ')',
+ '/', '(', '2', 'a', ')', ' ', ' ', 0xE6, 0x97, 0xA5, 0xE6, 0x9C, 0xAC,
+ 0xE8, 0xAA, 0x9E, ',', 0xE4, 0xB8, 0xAD, 0xE6, 0x96, 0x87, ',',
+ 0xE1, 0x84, 0x92, 0xE1, 0x85, 0xA1, 0xE1, 0x86, 0xAB,
+ 0xE1, 0x84, 0x80, 0xE1, 0x85, 0xB3, 0xE1, 0x86, 0xAF, '\n'
+ };
+ ASSERT (check (input, SIZEOF (input), expected, SIZEOF (expected)) == 0);
+ }
+
+#if HAVE_DECL_ALARM
+ /* Declare failure if test takes too long, by using default abort
+ caused by SIGALRM. */
+ signal (SIGALRM, SIG_DFL);
+ alarm (50);
+#endif
+
+ /* Check that the sorting is not O(n²) but O(n log n). */
+ {
+ int pass;
+ for (pass = 0; pass < 3; pass++)
+ {
+ size_t repeat = 1;
+ size_t m = 100000;
+ uint8_t *input = (uint8_t *) malloc (2 * (2 * m - 1) * sizeof (uint8_t));
+ if (input != NULL)
+ {
+ uint8_t *expected = input + (2 * m - 1);
+ size_t m1 = m / 2;
+ size_t m2 = (m - 1) / 2;
+ /* NB: m1 + m2 == m - 1. */
+ uint8_t *p;
+ size_t i;
+
+ input[0] = 0x41;
+ p = input + 1;
+ switch (pass)
+ {
+ case 0:
+ for (i = 0; i < m1; i++)
+ {
+ *p++ = 0xCC;
+ *p++ = 0x99;
+ }
+ for (i = 0; i < m2; i++)
+ {
+ *p++ = 0xCC;
+ *p++ = 0x80;
+ }
+ break;
+
+ case 1:
+ for (i = 0; i < m2; i++)
+ {
+ *p++ = 0xCC;
+ *p++ = 0x80;
+ }
+ for (i = 0; i < m1; i++)
+ {
+ *p++ = 0xCC;
+ *p++ = 0x99;
+ }
+ break;
+
+ case 2:
+ for (i = 0; i < m2; i++)
+ {
+ *p++ = 0xCC;
+ *p++ = 0x99;
+ *p++ = 0xCC;
+ *p++ = 0x80;
+ }
+ for (; i < m1; i++)
+ {
+ *p++ = 0xCC;
+ *p++ = 0x99;
+ }
+ break;
+
+ default:
+ abort ();
+ }
+
+ expected[0] = 0x41;
+ p = expected + 1;
+ for (i = 0; i < m1; i++)
+ {
+ *p++ = 0xCC;
+ *p++ = 0x99;
+ }
+ for (i = 0; i < m2; i++)
+ {
+ *p++ = 0xCC;
+ *p++ = 0x80;
+ }
+
+ for (; repeat > 0; repeat--)
+ ASSERT (check (input, 2 * m - 1, expected, 2 * m - 1) == 0);
+
+ free (input);
+ }
+ }
+ }
+}
+
+#else
+
+void
+test_u8_nfd (void)
+{
+}
+
+#endif