std::vector<lstring> convert_split_words(const lstring <) {
std::vector<lstring> ret;
UBreakIterator* bi;
int prev = -1, pos;
UErrorCode err = U_ZERO_ERROR;
bi = ubrk_open(UBRK_WORD, get_locale(), (UChar *)lt.data(), lt.size(), &err);
if (U_FAILURE(err))
return ret;
pos = ubrk_first(bi);
while (pos != UBRK_DONE) {
int rules = ubrk_getRuleStatus(bi);
if ((rules == UBRK_WORD_NONE) || (prev == -1)) {
prev = pos;
} else {
ret.emplace_back(lt.substr(prev, pos - prev));
prev = -1;
}
pos = ubrk_next(bi);
}
ubrk_close(bi);
return ret;
}
// BreakIterator.split {{{
static PyObject *
icu_BreakIterator_split(icu_BreakIterator *self, PyObject *args, PyObject *kwargs) {
int32_t prev = 0, p = 0, sz = 0;
PyObject *ans = NULL, *token = NULL;
ans = PyList_New(0);
if (ans == NULL) return PyErr_NoMemory();
p = ubrk_first(self->break_iterator);
while (p != UBRK_DONE) {
prev = p; p = ubrk_next(self->break_iterator);
if (self->type == UBRK_WORD && ubrk_getRuleStatus(self->break_iterator) == UBRK_WORD_NONE)
continue; // We are not at the start of a word
sz = (p == UBRK_DONE) ? self->text_len - prev : p - prev;
if (sz > 0) {
token = icu_to_python(self->text + prev, sz);
if (token == NULL) {
Py_DECREF(ans); ans = NULL; break;
}
if (PyList_Append(ans, token) != 0) {
Py_DECREF(token); Py_DECREF(ans); ans = NULL; break;
}
Py_DECREF(token);
}
}
return ans;
} // }}}
// BreakIterator.index {{{
static PyObject *
icu_BreakIterator_index(icu_BreakIterator *self, PyObject *token) {
#if PY_VERSION_HEX >= 0x03030000
#error Not implemented for python >= 3.3
#endif
UChar *buf = NULL, *needle = NULL;
int32_t word_start = 0, p = 0, sz = 0, ans = -1, leading_hyphen = 0, trailing_hyphen = 0;
buf = python_to_icu(token, &sz);
if (buf == NULL) return NULL;
if (sz < 1) goto end;
needle = buf;
if (sz > 1 && IS_HYPHEN_CHAR(buf[0])) { needle = buf + 1; leading_hyphen = 1; sz -= 1; }
if (sz > 1 && IS_HYPHEN_CHAR(buf[sz-1])) trailing_hyphen = 1;
Py_BEGIN_ALLOW_THREADS;
p = ubrk_first(self->break_iterator);
while (p != UBRK_DONE) {
word_start = p; p = ubrk_next(self->break_iterator);
if (self->type == UBRK_WORD && ubrk_getRuleStatus(self->break_iterator) == UBRK_WORD_NONE)
continue; // We are not at the start of a word
if (self->text_len >= word_start + sz && memcmp(self->text + word_start, needle, sz * sizeof(UChar)) == 0) {
if (word_start > 0 && (
(leading_hyphen && !IS_HYPHEN_CHAR(self->text[word_start-1])) ||
(!leading_hyphen && IS_HYPHEN_CHAR(self->text[word_start-1]))
)) continue;
if (!trailing_hyphen && IS_HYPHEN_CHAR(self->text[word_start + sz])) continue;
if (p == UBRK_DONE || self->text_len <= word_start + sz) { ans = word_start; break; }
if (
// Check that the found word is followed by a word boundary
ubrk_isBoundary(self->break_iterator, word_start + sz) &&
// If there is a leading hyphen check that the leading
// hyphen is preceded by a word boundary
(!leading_hyphen || (word_start > 1 && ubrk_isBoundary(self->break_iterator, word_start - 2))) &&
// Check that there is a word boundary *after* the trailing
// hyphen. We cannot rely on ubrk_isBoundary() as that
// always returns true because of the trailing hyphen.
(!trailing_hyphen || ubrk_following(self->break_iterator, word_start + sz) == UBRK_DONE || ubrk_getRuleStatus(self->break_iterator) == UBRK_WORD_NONE)
) { ans = word_start; break; }
if (p != UBRK_DONE) ubrk_isBoundary(self->break_iterator, p); // Reset the iterator to its position before the call to ubrk_isBoundary()
}
}
if (leading_hyphen && ans > -1) ans -= 1;
#ifdef Py_UNICODE_WIDE
if (ans > 0) ans = u_countChar32(self->text, ans);
#endif
Py_END_ALLOW_THREADS;
end:
free(buf);
return Py_BuildValue("l", (long)ans);
} // }}}
/**
* Is current element in this iterator valid?
* If result is false, the element will be skipped.
* Used for word_only.
*/
inline int is_valid_elem(cloner* res, UBreakIterator* iter)
{
cloner_break* res_brk = (cloner_break*) res;
int32_t type;
if (res_brk->skip_elem == NULL)
return 1;
type = ubrk_getRuleStatus(iter);
return !((res_brk->skip_elem)(type));
}
/*
* TestBreakIteratorRules - Verify that a break iterator can be created from
* a set of source rules.
*/
static void TestBreakIteratorRules() {
/* Rules will keep together any run of letters not including 'a', OR
* keep together 'abc', but only when followed by 'def', OTHERWISE
* just return one char at a time.
*/
char rules[] = "abc{666}/def;\n [\\p{L} - [a]]* {2}; . {1};";
/* 0123456789012345678 */
char data[] = "abcdex abcdefgh-def"; /* the test data string */
char breaks[] = "** ** * ** *"; /* * the expected break positions */
char tags[] = "01 21 6 21 2"; /* expected tag values at break positions */
int32_t tagMap[] = {0, 1, 2, 3, 4, 5, 666};
UChar *uData;
void *freeHook = NULL;
UErrorCode status = U_ZERO_ERROR;
int32_t pos;
int i;
UBreakIterator *bi = testOpenRules(rules);
if (bi == NULL) {return;}
uData = toUChar(data, &freeHook);
ubrk_setText(bi, uData, -1, &status);
pos = ubrk_first(bi);
for (i=0; i<sizeof(breaks); i++) {
if (pos == i && breaks[i] != '*') {
log_err("FAIL: unexpected break at position %d found\n", pos);
break;
}
if (pos != i && breaks[i] == '*') {
log_err("FAIL: expected break at position %d not found.\n", i);
break;
}
if (pos == i) {
int32_t tag, expectedTag;
tag = ubrk_getRuleStatus(bi);
expectedTag = tagMap[tags[i]&0xf];
if (tag != expectedTag) {
log_err("FAIL: incorrect tag value. Position = %d; expected tag %d, got %d",
pos, expectedTag, tag);
break;
}
pos = ubrk_next(bi);
}
}
freeToUCharStrings(&freeHook);
ubrk_close(bi);
}
MojErr MojDbTextTokenizer::tokenize(const MojString& text, MojDbTextCollator* collator, KeySet& keysOut) const
{
LOG_TRACE("Entering function %s", __FUNCTION__);
MojAssert(m_ubrk.get());
// convert to UChar from str
MojDbTextUtils::UnicodeVec unicodeStr;
MojErr err = MojDbTextUtils::strToUnicode(text, unicodeStr);
MojErrCheck(err);
// clone break iterator and set text
MojByte buf[U_BRK_SAFECLONE_BUFFERSIZE];
UErrorCode status = U_ZERO_ERROR;
MojInt32 size = sizeof(buf);
IterPtr ubrk(ubrk_safeClone(m_ubrk.get(), buf, &size, &status));
MojUnicodeErrCheck(status);
MojAssert(ubrk.get());
ubrk_setText(ubrk.get(), unicodeStr.begin(), (MojInt32) unicodeStr.size(), &status);
MojUnicodeErrCheck(status);
MojInt32 tokBegin = -1;
MojInt32 pos = ubrk_first(ubrk.get());
while (pos != UBRK_DONE) {
UWordBreak status = (UWordBreak) ubrk_getRuleStatus(ubrk.get());
if (status != UBRK_WORD_NONE) {
MojAssert(tokBegin != -1);
MojDbKey key;
const UChar* tokChars = unicodeStr.begin() + tokBegin;
MojSize tokSize = (MojSize) (pos - tokBegin);
if (collator) {
err = collator->sortKey(tokChars, tokSize, key);
MojErrCheck(err);
} else {
MojString tok;
err = MojDbTextUtils::unicodeToStr(tokChars, tokSize, tok);
MojErrCheck(err);
err = key.assign(tok);
MojErrCheck(err);
}
err = keysOut.put(key);
MojErrCheck(err);
}
tokBegin = pos;
pos = ubrk_next(ubrk.get());
}
return MojErrNone;
}
// BreakIterator.index {{{
static PyObject *
icu_BreakIterator_index(icu_BreakIterator *self, PyObject *args, PyObject *kwargs) {
#if PY_VERSION_HEX >= 0x03030000
#error Not implemented for python >= 3.3
#endif
UChar *buf = NULL;
int32_t prev = 0, p = 0, sz = 0, tsz = 0, ans = -1;
PyObject *token = NULL;
if (!PyArg_ParseTuple(args, "O", &token)) return NULL;
buf = python_to_icu(token, &sz, 1);
if (buf == NULL) return NULL;
if (sz < 1) goto end;
Py_BEGIN_ALLOW_THREADS;
p = ubrk_first(self->break_iterator);
while (p != UBRK_DONE) {
prev = p; p = ubrk_next(self->break_iterator);
if (self->type == UBRK_WORD && ubrk_getRuleStatus(self->break_iterator) == UBRK_WORD_NONE)
continue; // We are not at the start of a word
tsz = (p == UBRK_DONE) ? self->text_len - prev : p - prev;
if (sz == tsz && memcmp(self->text + prev, buf, sz * sizeof(UChar)) == 0) {
#ifdef PY_UNICODE_WIDE
ans = u_countChar32(self->text, prev);
#else
ans = prev;
#endif
break;
}
}
Py_END_ALLOW_THREADS;
end:
free(buf);
return Py_BuildValue("i", ans);
} // }}}
// BreakIterator.split2 {{{
static PyObject *
icu_BreakIterator_split2(icu_BreakIterator *self, PyObject *args, PyObject *kwargs) {
#if PY_VERSION_HEX >= 0x03030000
#error Not implemented for python >= 3.3
#endif
int32_t prev = 0, p = 0, sz = 0;
PyObject *ans = NULL, *temp = NULL;
ans = PyList_New(0);
if (ans == NULL) return PyErr_NoMemory();
p = ubrk_first(self->break_iterator);
while (p != UBRK_DONE) {
prev = p; p = ubrk_next(self->break_iterator);
if (self->type == UBRK_WORD && ubrk_getRuleStatus(self->break_iterator) == UBRK_WORD_NONE)
continue; // We are not at the start of a word
sz = (p == UBRK_DONE) ? self->text_len - prev : p - prev;
if (sz > 0) {
#ifdef Py_UNICODE_WIDE
sz = u_countChar32(self->text + prev, sz);
prev = u_countChar32(self->text, prev);
#endif
temp = Py_BuildValue("II", prev, sz);
if (temp == NULL) {
Py_DECREF(ans); ans = NULL; break;
}
if (PyList_Append(ans, temp) != 0) {
Py_DECREF(temp); Py_DECREF(ans); ans = NULL; break;
}
Py_DECREF(temp);
}
}
return ans;
} // }}}
// BreakIterator.split2 {{{
static PyObject *
icu_BreakIterator_split2(icu_BreakIterator *self, PyObject *args) {
#if PY_VERSION_HEX >= 0x03030000
#error Not implemented for python >= 3.3
#endif
int32_t word_start = 0, p = 0, sz = 0, last_pos = 0, last_sz = 0;
int is_hyphen_sep = 0, leading_hyphen = 0, trailing_hyphen = 0;
UChar sep = 0;
PyObject *ans = NULL, *temp = NULL, *t = NULL;
ans = PyList_New(0);
if (ans == NULL) return PyErr_NoMemory();
p = ubrk_first(self->break_iterator);
while (p != UBRK_DONE) {
word_start = p; p = ubrk_next(self->break_iterator);
if (self->type == UBRK_WORD && ubrk_getRuleStatus(self->break_iterator) == UBRK_WORD_NONE)
continue; // We are not at the start of a word
sz = (p == UBRK_DONE) ? self->text_len - word_start : p - word_start;
if (sz > 0) {
// ICU breaks on words containing hyphens, we do not want that, so we recombine manually
is_hyphen_sep = 0; leading_hyphen = 0; trailing_hyphen = 0;
if (word_start > 0) { // Look for a leading hyphen
sep = *(self->text + word_start - 1);
if (IS_HYPHEN_CHAR(sep)) {
leading_hyphen = 1;
if (last_pos > 0 && word_start - last_pos == 1) is_hyphen_sep = 1;
}
}
if (word_start + sz < self->text_len) { // Look for a trailing hyphen
sep = *(self->text + word_start + sz);
if (IS_HYPHEN_CHAR(sep)) trailing_hyphen = 1;
}
last_pos = p;
#ifdef Py_UNICODE_WIDE
sz = u_countChar32(self->text + word_start, sz);
word_start = u_countChar32(self->text, word_start);
#endif
if (is_hyphen_sep && PyList_GET_SIZE(ans) > 0) {
sz = last_sz + sz + trailing_hyphen;
last_sz = sz;
t = PyInt_FromLong((long)sz);
if (t == NULL) { Py_DECREF(ans); ans = NULL; break; }
temp = PyList_GET_ITEM(ans, PyList_GET_SIZE(ans) - 1);
Py_DECREF(PyTuple_GET_ITEM(temp, 1));
PyTuple_SET_ITEM(temp, 1, t);
} else {
sz += leading_hyphen + trailing_hyphen;
last_sz = sz;
temp = Py_BuildValue("ll", (long)(word_start - leading_hyphen), (long)sz);
if (temp == NULL) {
Py_DECREF(ans); ans = NULL; break;
}
if (PyList_Append(ans, temp) != 0) {
Py_DECREF(temp); Py_DECREF(ans); ans = NULL; break;
}
Py_DECREF(temp);
}
}
}
return ans;
} // }}}
bool isWordTextBreak(TextBreakIterator* iterator)
{
int ruleStatus = ubrk_getRuleStatus(reinterpret_cast<UBreakIterator*>(iterator));
return ruleStatus != UBRK_WORD_NONE;
}
bool isWordTextBreak(UBreakIterator* iterator)
{
int ruleStatus = ubrk_getRuleStatus(iterator);
return ruleStatus != UBRK_WORD_NONE;
}
static engine_return_t engine_fixed_match(error_t **error, void *data, const UString *subject)
{
int32_t ret;
UErrorCode status;
FETCH_DATA(data, p, fixed_pattern_t);
status = U_ZERO_ERROR;
if (ustring_empty(p->pattern)) {
if (IS_WORD_BOUNDED(p->flags)) {
if (ustring_empty(subject)) {
return ENGINE_MATCH_FOUND;
} else {
int32_t l, u, lastState, state;
ubrk_setText(p->ubrk, subject->ptr, subject->len, &status);
if (U_FAILURE(status)) {
icu_error_set(error, FATAL, status, "ubrk_setText");
return ENGINE_FAILURE;
}
if (UBRK_DONE != (l = ubrk_first(p->ubrk))) {
lastState = ubrk_getRuleStatus(p->ubrk);
while (UBRK_DONE != (u = ubrk_next(p->ubrk))) {
state = ubrk_getRuleStatus(p->ubrk);
if (UBRK_WORD_NONE == lastState && lastState == state) {
return ENGINE_MATCH_FOUND;
}
lastState = state;
l = u;
}
}
return ENGINE_NO_MATCH;
}
} else {
return ENGINE_MATCH_FOUND;
}
} else if (NULL != p->usearch) {
if (subject->len > 0) {
usearch_setText(p->usearch, subject->ptr, subject->len, &status);
if (U_FAILURE(status)) {
icu_error_set(error, FATAL, status, "usearch_setText");
return ENGINE_FAILURE;
}
ret = usearch_first(p->usearch, &status);
if (U_FAILURE(status)) {
icu_error_set(error, FATAL, status, "usearch_first");
return ENGINE_FAILURE;
}
usearch_unbindText(p->usearch);
return (ret != USEARCH_DONE ? ENGINE_MATCH_FOUND : ENGINE_NO_MATCH);
} else {
return ENGINE_NO_MATCH;
}
} else {
UChar *m;
int32_t pos;
pos = 0;
ret = ENGINE_NO_MATCH;
if (NULL != p->ubrk) {
ubrk_setText(p->ubrk, subject->ptr, subject->len, &status);
if (U_FAILURE(status)) {
icu_error_set(error, FATAL, status, "ubrk_setText");
return ENGINE_FAILURE;
}
}
while (NULL != (m = u_strFindFirst(subject->ptr + pos, subject->len - pos, p->pattern->ptr, p->pattern->len))) {
pos = m - subject->ptr;
if (NULL == p->ubrk || (ubrk_isBoundary(p->ubrk, pos) && ubrk_isBoundary(p->ubrk, pos + p->pattern->len))) {
ret = ENGINE_MATCH_FOUND;
}
pos += p->pattern->len;
}
ubrk_unbindText(p->ubrk);
return ret;
}
}
static engine_return_t engine_fixed_match_all(error_t **error, void *data, const UString *subject, interval_list_t *intervals)
{
int32_t matches;
UErrorCode status;
FETCH_DATA(data, p, fixed_pattern_t);
matches = 0;
status = U_ZERO_ERROR;
if (ustring_empty(p->pattern)) {
if (IS_WORD_BOUNDED(p->flags)) {
if (ustring_empty(subject)) {
return ENGINE_MATCH_FOUND;
} else {
int32_t l, u, lastState, state;
ubrk_setText(p->ubrk, subject->ptr, subject->len, &status);
if (U_FAILURE(status)) {
icu_error_set(error, FATAL, status, "ubrk_setText");
return ENGINE_FAILURE;
}
if (UBRK_DONE != (l = ubrk_first(p->ubrk))) {
lastState = ubrk_getRuleStatus(p->ubrk);
while (UBRK_DONE != (u = ubrk_next(p->ubrk))) {
state = ubrk_getRuleStatus(p->ubrk);
if (UBRK_WORD_NONE == lastState && lastState == state) {
return ENGINE_MATCH_FOUND;
}
lastState = state;
l = u;
}
}
return ENGINE_NO_MATCH;
}
} else {
return ENGINE_MATCH_FOUND;
}
} else if (NULL != p->usearch) {
int32_t l, u;
if (subject->len > 0) {
usearch_setText(p->usearch, subject->ptr, subject->len, &status);
if (U_FAILURE(status)) {
icu_error_set(error, FATAL, status, "usearch_setText");
return ENGINE_FAILURE;
}
for (l = usearch_first(p->usearch, &status); U_SUCCESS(status) && USEARCH_DONE != l; l = usearch_next(p->usearch, &status)) {
matches++;
u = l + usearch_getMatchedLength(p->usearch);
if (interval_list_add(intervals, subject->len, l, u)) {
return ENGINE_WHOLE_LINE_MATCH;
}
}
if (U_FAILURE(status)) {
icu_error_set(error, FATAL, status, "usearch_[first|next]");
return ENGINE_FAILURE;
}
usearch_unbindText(p->usearch);
return (matches ? ENGINE_MATCH_FOUND : ENGINE_NO_MATCH);
} else {
return ENGINE_NO_MATCH;
}
} else {
UChar *m;
int32_t pos;
pos = 0;
if (NULL != p->ubrk) {
ubrk_setText(p->ubrk, subject->ptr, subject->len, &status);
if (U_FAILURE(status)) {
icu_error_set(error, FATAL, status, "ubrk_setText");
return ENGINE_FAILURE;
}
}
while (NULL != (m = u_strFindFirst(subject->ptr + pos, subject->len - pos, p->pattern->ptr, p->pattern->len))) {
pos = m - subject->ptr;
if (NULL == p->ubrk || (ubrk_isBoundary(p->ubrk, pos) && ubrk_isBoundary(p->ubrk, pos + p->pattern->len))) {
matches++;
if (interval_list_add(intervals, subject->len, pos, pos + p->pattern->len)) {
return ENGINE_WHOLE_LINE_MATCH;
}
}
pos += p->pattern->len;
}
ubrk_unbindText(p->ubrk);
return (matches ? ENGINE_MATCH_FOUND : ENGINE_NO_MATCH);
}
}
int icu_breakpoints(lua_State *L) {
const char* input = luaL_checkstring(L, 1);
int input_l = strlen(input);
const char* locale = luaL_checkstring(L, 2);
UChar *buffer;
int32_t l, breakcount = 0;
UErrorCode err = U_ZERO_ERROR;
u_strFromUTF8(NULL, 0, &l, input, input_l, &err);
/* Above call returns an error every time. */
err = U_ZERO_ERROR;
buffer = malloc(l * sizeof(UChar));
u_strFromUTF8(buffer, l, &l, input, input_l, &err);
UBreakIterator* wordbreaks, *linebreaks;
int32_t i, previous;
wordbreaks = ubrk_open(UBRK_WORD, locale, buffer, l, &err);
if(U_FAILURE(err)) {
luaL_error(L, "Word break parser failure: %s", u_errorName(err));
}
linebreaks = ubrk_open(UBRK_LINE, locale, buffer, l, &err);
if(U_FAILURE(err)) {
luaL_error(L, "Line break parser failure: %s", u_errorName(err));
}
previous = 0;
i = 0;
while (i <= l) {
int32_t out_l;
int32_t type;
if (!ubrk_isBoundary(linebreaks, i) && !ubrk_isBoundary(wordbreaks,i)) {
i++; continue;
}
lua_checkstack(L, 3);
/* At some kind of boundary */
lua_newtable(L);
lua_pushstring(L, "type");
lua_pushstring(L, ubrk_isBoundary(linebreaks,i) ? "line" : "word");
lua_settable(L, -3);
int32_t utf8_index = 0;
err = U_ZERO_ERROR;
u_strToUTF8(NULL, 0, &utf8_index, buffer, i, &err);
assert(U_SUCCESS(err) || err == U_BUFFER_OVERFLOW_ERROR);
lua_pushstring(L, "index");
lua_pushinteger(L, utf8_index);
lua_settable(L, -3);
if (ubrk_isBoundary(linebreaks, i)) {
lua_pushstring(L, "subtype");
type = ubrk_getRuleStatus(linebreaks);
if (type >= UBRK_LINE_SOFT && type < UBRK_LINE_SOFT_LIMIT) {
lua_pushstring(L, "soft");
} else {
lua_pushstring(L, "hard");
}
lua_settable(L, -3);
}
lua_pushstring(L, "token");
lua_pushlstring(L, input+previous, utf8_index-previous);
lua_settable(L, -3);
previous = utf8_index;
breakcount++;
i++;
}
ubrk_close(wordbreaks);
ubrk_close(linebreaks);
return breakcount;
}
int32_t __hs_ubrk_getRuleStatus(UBreakIterator *bi)
{
return ubrk_getRuleStatus(bi);
}
