1 /*
2 * Copyright (C) 2017 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 * http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17 #include "smartselect/tokenizer.h"
18
19 #include "util/strings/utf8.h"
20 #include "util/utf8/unicodetext.h"
21
22 namespace libtextclassifier {
23
PrepareTokenizationCodepointRanges(const std::vector<TokenizationCodepointRange> & codepoint_range_configs)24 void Tokenizer::PrepareTokenizationCodepointRanges(
25 const std::vector<TokenizationCodepointRange>& codepoint_range_configs) {
26 codepoint_ranges_.clear();
27 codepoint_ranges_.reserve(codepoint_range_configs.size());
28 for (const TokenizationCodepointRange& range : codepoint_range_configs) {
29 codepoint_ranges_.push_back(
30 CodepointRange(range.start(), range.end(), range.role()));
31 }
32
33 std::sort(codepoint_ranges_.begin(), codepoint_ranges_.end(),
34 [](const CodepointRange& a, const CodepointRange& b) {
35 return a.start < b.start;
36 });
37 }
38
FindTokenizationRole(int codepoint) const39 TokenizationCodepointRange::Role Tokenizer::FindTokenizationRole(
40 int codepoint) const {
41 auto it = std::lower_bound(codepoint_ranges_.begin(), codepoint_ranges_.end(),
42 codepoint,
43 [](const CodepointRange& range, int codepoint) {
44 // This function compares range with the
45 // codepoint for the purpose of finding the first
46 // greater or equal range. Because of the use of
47 // std::lower_bound it needs to return true when
48 // range < codepoint; the first time it will
49 // return false the lower bound is found and
50 // returned.
51 //
52 // It might seem weird that the condition is
53 // range.end <= codepoint here but when codepoint
54 // == range.end it means it's actually just
55 // outside of the range, thus the range is less
56 // than the codepoint.
57 return range.end <= codepoint;
58 });
59 if (it != codepoint_ranges_.end() && it->start <= codepoint &&
60 it->end > codepoint) {
61 return it->role;
62 } else {
63 return TokenizationCodepointRange::DEFAULT_ROLE;
64 }
65 }
66
Tokenize(const std::string & utf8_text) const67 std::vector<Token> Tokenizer::Tokenize(const std::string& utf8_text) const {
68 UnicodeText context_unicode = UTF8ToUnicodeText(utf8_text, /*do_copy=*/false);
69
70 std::vector<Token> result;
71 Token new_token("", 0, 0);
72 int codepoint_index = 0;
73 for (auto it = context_unicode.begin(); it != context_unicode.end();
74 ++it, ++codepoint_index) {
75 TokenizationCodepointRange::Role role = FindTokenizationRole(*it);
76 if (role & TokenizationCodepointRange::SPLIT_BEFORE) {
77 if (!new_token.value.empty()) {
78 result.push_back(new_token);
79 }
80 new_token = Token("", codepoint_index, codepoint_index);
81 }
82 if (!(role & TokenizationCodepointRange::DISCARD_CODEPOINT)) {
83 new_token.value += std::string(
84 it.utf8_data(),
85 it.utf8_data() + GetNumBytesForNonZeroUTF8Char(it.utf8_data()));
86 ++new_token.end;
87 }
88 if (role & TokenizationCodepointRange::SPLIT_AFTER) {
89 if (!new_token.value.empty()) {
90 result.push_back(new_token);
91 }
92 new_token = Token("", codepoint_index + 1, codepoint_index + 1);
93 }
94 }
95 if (!new_token.value.empty()) {
96 result.push_back(new_token);
97 }
98
99 return result;
100 }
101
102 } // namespace libtextclassifier
103