diff options
Diffstat (limited to 'text/test-text.c')
| -rw-r--r-- | text/test-text.c | 79 |
1 files changed, 77 insertions, 2 deletions
diff --git a/text/test-text.c b/text/test-text.c index e8077b5..a367330 100644 --- a/text/test-text.c +++ b/text/test-text.c @@ -127,13 +127,88 @@ static size_t bidi_character_test(const char *path) { return cases; } +/* NormalizationTest.txt (through bzcat): for each line c1;c2;c3;c4;c5, + * NFC(c1) = NFC(c2) = NFC(c3) = c2, NFC(c4) = NFC(c5) = c4, NFD(c1) = + * NFD(c2) = NFD(c3) = c3 and NFD(c4) = NFD(c5) = c5; and every code point + * part 1 does not list is its own NFC and NFD. */ +static size_t parse_codes(const char *text, uint32_t *codes) { + size_t n = 0; + for (char *end; *text && n < MAX; text = end) { + unsigned long code = strtoul(text, &end, 16); + if (end == text) break; + codes[n++] = (uint32_t)code; + } + return n; +} + +static bool normalizes(size_t (*form)(const uint32_t *, size_t, uint32_t *), const uint32_t *in, size_t count, + const uint32_t *want, size_t want_count) { + uint32_t out[4 * MAX]; + size_t n = form(in, count, out); + return n == want_count && !memcmp(out, want, n * sizeof(*out)); +} + +static size_t normalization_test(const char *path, size_t *invariant) { + char command[1200]; + snprintf(command, sizeof(command), "bzcat '%s'", path); + FILE *in = popen(command, "r"); + if (!in) { perror(path); exit(2); } + static bool listed[0x110000]; + char line[4096]; + size_t cases = 0, number = 0; + int part = -1; + while (fgets(line, sizeof(line), in)) { + ++number; + if (line[0] == '#' || line[0] == '\n') continue; + if (!strncmp(line, "@Part", 5)) { part = atoi(line + 5); continue; } + uint32_t c[5][MAX]; + size_t n[5]; + char *field = line; + for (unsigned f = 0; f < 5; ++f) { + char *end = strchr(field, ';'); + if (!end) break; + *end = 0; + n[f] = parse_codes(field, c[f]); + field = end + 1; + } + if (part == 1 && n[0] == 1) listed[c[0][0]] = true; + ++cases; + bool ok = true; + for (unsigned f = 0; f < 3; ++f) { + ok = ok && normalizes(vega_nfc, c[f], n[f], c[1], n[1]); + ok = ok && normalizes(vega_nfd, c[f], n[f], c[2], n[2]); + } + for (unsigned f = 3; f < 5; ++f) { + ok = ok && normalizes(vega_nfc, c[f], n[f], c[3], n[3]); + ok = ok && normalizes(vega_nfd, c[f], n[f], c[4], n[4]); + } + if (!ok) fail(path, number, "a form differs"); + } + pclose(in); + *invariant = 0; + for (uint32_t code = 0; code < 0x110000; ++code) { + if (listed[code] || (code >= 0xD800 && code <= 0xDFFF)) continue; + ++*invariant; + if (!normalizes(vega_nfc, &code, 1, &code, 1) || !normalizes(vega_nfd, &code, 1, &code, 1)) { + char what[64]; + snprintf(what, sizeof(what), "U+%04X is not its own normal form", code); + fail(path, 0, what); + } + } + return cases; +} + int main(int argc, char **argv) { const char *ucd = argc > 1 ? argv[1] : "/usr/share/unicode"; char path[1024]; snprintf(path, sizeof(path), "%s/BidiTest.txt", ucd); size_t bidi = bidi_test(path); snprintf(path, sizeof(path), "%s/BidiCharacterTest.txt", ucd); - size_t characters = bidi_character_test(path); - printf("{\"bidi_test_cases\":%zu,\"bidi_character_test_cases\":%zu,\"failures\":%u}\n", bidi, characters, failures); + size_t characters = bidi_character_test(path), invariant; + snprintf(path, sizeof(path), "%s/NormalizationTest.txt.bz2", ucd); + size_t normalized = normalization_test(path, &invariant); + printf("{\"bidi_test_cases\":%zu,\"bidi_character_test_cases\":%zu,\"normalization_test_cases\":%zu," + "\"normalization_invariant_code_points\":%zu,\"failures\":%u}\n", bidi, characters, normalized, invariant, + failures); return failures ? 1 : 0; } |