1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155
|
/* Copyright (C) 2022 J.F.Dockes
*
* License: GPL 2.1
*
* 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 2.1 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 Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public License
* along with this program; if not, write to the
* Free Software Foundation, Inc.,
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
*/
#include "cjksplitter.h"
#include <cassert>
#include "smallut.h"
#include "log.h"
#include "utf8iter.h"
static const int o_CJKMaxNgramLen{5};
int CJKSplitter::max_ngramlen()
{
return o_CJKMaxNgramLen;
}
// Note that this should probably be renamed NGRAMSplitter as this is also used for other scripts.
//
// We output ngrams.
// For example: for input "abc" and ngramlen==2, we generate: a ab b bc c as words.
//
// This is very different from the normal behaviour, so we don't use doemit() or emitterm().
//
// The routine is sort of a mess and goes to show that we'd probably be better off converting the
// whole buffer to utf32 on entry...
bool CJKSplitter::text_to_words(Utf8Iter& it, unsigned int *cp, int& wordpos)
{
LOGDEB1("cjk_to_words: wordpos " << wordpos << "\n");
int flags = m_sink.flags();
// We use an offset buffer to remember the starts of the utf-8
// characters which we still need to use.
assert(m_ngramlen < o_CJKMaxNgramLen);
std::string::size_type boffs[o_CJKMaxNgramLen+1];
std::string mybuf;
std::string::size_type myboffs[o_CJKMaxNgramLen+1];
// Current number of valid offsets;
int nchars = 0;
unsigned int c = 0;
bool spacebefore{false};
for (; !it.eof() && !it.error(); it++) {
c = *it;
LOGDEB1("Processing [" << (std::string)it << "]\n");
// We had a version which ignored whitespace for some time,
// but this was a bad idea. Only break on a non-cjk
// *alphabetic* character, except if following punctuation, in
// which case we return for any non-cjk. This allows compound
// cjk+numeric spans, or punctuated cjk spans to be
// continually indexed as cjk. The best approach is a matter
// of appreciation...
if ((spacebefore || (c > 255
#ifdef TESTING_NGRAMS
|| islower(c)
#else
|| isalpha(c)
#endif
)) && !TextSplit::isNGRAMMED(c)) {
// Return to normal handler
break;
}
if (TextSplit::isSpace(c)) {
// Flush the ngram buffer and go on
nchars = 0;
mybuf.clear();
spacebefore = true;
continue;
} else {
spacebefore = false;
}
if (nchars == m_ngramlen) {
// Offset buffer full, shift it. Might be more efficient
// to have a circular one, but things are complicated
// enough already...
for (int i = 0; i < nchars-1; i++) {
boffs[i] = boffs[i+1];
}
for (int i = 0; i < nchars-1; i++) {
myboffs[i] = myboffs[i+1];
}
} else {
nchars++;
}
// Copy to local buffer, and note local offset
myboffs[nchars-1] = mybuf.size();
it.appendchartostring(mybuf);
// Take note of document byte offset for this character.
boffs[nchars-1] = it.getBpos();
// Output all new ngrams: they begin at each existing position
// and end after the new character. onlyspans->only output
// maximum words, nospans=> single chars
if (!(flags & TextSplit::TXTS_ONLYSPANS) || nchars == m_ngramlen) {
auto btend = it.getBpos() + it.getBlen();
int loopbeg = (flags & TextSplit::TXTS_NOSPANS) ? nchars-1 : 0;
int loopend = (flags & TextSplit::TXTS_ONLYSPANS) ? 1 : nchars;
for (int i = loopbeg; i < loopend; i++) {
// Because of the whitespace handling above there may be whitespace in the
// buffer. Strip it from the output words. This means that the offs/size will be
// slightly off (->highlights), to be fixed one day.
auto word = mybuf.substr(myboffs[i], mybuf.size() - myboffs[i]);
LOGDEB1("EMITTING [" << trimstring(word, "\r\n\f \t") << "] at pos " <<
wordpos - (nchars - i - 1) << '\n');
if (!m_sink.takeword(trimstring(word, "\r\n\f \t"),
wordpos - (nchars - i - 1), boffs[i], btend)) {
return false;
}
}
if ((flags & TextSplit::TXTS_ONLYSPANS)) {
// Only spans: don't overlap: flush buffer
nchars = 0;
mybuf.clear();
}
}
// Increase word position by one, other words are at an
// existing position. This could be subject to discussion...
wordpos++;
}
// If onlyspans is set, there may be things to flush in the buffer
// first
if ((flags & TextSplit::TXTS_ONLYSPANS) && nchars > 0 && nchars != m_ngramlen) {
int btend = int(it.getBpos()); // Current char is out
// See comment before takeword above.
auto word = mybuf.substr(myboffs[0], mybuf.size() - myboffs[0]);
LOGDEB1("EMITTING [" << trimstring(word, "\r\n\f \t") << "] at pos " <<
wordpos - nchars << '\n');
if (!m_sink.takeword(trimstring(word, "\r\n\f \t"), wordpos - nchars, boffs[0], btend)) {
return false;
}
}
*cp = c;
return true;
}
|