1 // Copyright (C) 2011 The Libphonenumber Authors
3 // Licensed under the Apache License, Version 2.0 (the "License");
4 // you may not use this file except in compliance with the License.
5 // You may obtain a copy of the License at
7 // http://www.apache.org/licenses/LICENSE-2.0
9 // Unless required by applicable law or agreed to in writing, software
10 // distributed under the License is distributed on an "AS IS" BASIS,
11 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 // See the License for the specific language governing permissions and
13 // limitations under the License.
15 // Author: Lara Rennie
18 // Implementation of a stateful class that finds and extracts telephone numbers
21 #include "phonenumbers/phonenumbermatcher.h"
23 #ifndef USE_ICU_REGEXP
24 #error phonenumbermatcher depends on ICU (i.e. USE_ICU_REGEXP must be set)
25 #endif // USE_ICU_REGEXP
33 #include <unicode/uchar.h>
35 #include "base/logging.h"
36 #include "base/memory/scoped_ptr.h"
37 #include "base/memory/singleton.h"
38 #include "phonenumbers/default_logger.h"
39 #include "phonenumbers/encoding_utils.h"
40 #include "phonenumbers/normalize_utf8.h"
41 #include "phonenumbers/phonemetadata.pb.h"
42 #include "phonenumbers/phonenumber.pb.h"
43 #include "phonenumbers/phonenumbermatch.h"
44 #include "phonenumbers/phonenumberutil.h"
45 #include "phonenumbers/regexp_adapter.h"
46 #include "phonenumbers/regexp_adapter_icu.h"
47 #include "phonenumbers/stringutil.h"
50 #include "phonenumbers/regexp_adapter_re2.h"
51 #endif // USE_RE2_AND_ICU
53 using std::numeric_limits;
58 namespace phonenumbers {
61 // Returns a regular expression quantifier with an upper and lower limit.
62 string Limit(int lower, int upper) {
65 DCHECK_LT(lower, upper);
66 return StrCat("{", lower, ",", upper, "}");
69 bool IsInvalidPunctuationSymbol(char32 character) {
70 return character == '%' || u_charType(character) == U_CURRENCY_SYMBOL;
73 // Helper method to get the national-number part of a number, formatted without
74 // any national prefix, and return it as a set of digit blocks that would be
75 // formatted together.
76 void GetNationalNumberGroups(const PhoneNumberUtil& util,
77 const PhoneNumber& number,
78 vector<string>* digit_blocks) {
79 // This will be in the format +CC-DG;ext=EXT where DG represents groups of
81 string rfc3966_format;
82 util.Format(number, PhoneNumberUtil::RFC3966, &rfc3966_format);
83 // We remove the extension part from the formatted string before splitting it
84 // into different groups.
85 size_t end_index = rfc3966_format.find(';');
86 if (end_index == string::npos) {
87 end_index = rfc3966_format.length();
89 // The country-code will have a '-' following it.
90 size_t start_index = rfc3966_format.find('-') + 1;
91 SplitStringUsing(rfc3966_format.substr(start_index, end_index - start_index),
95 bool ContainsOnlyValidXChars(const PhoneNumber& number, const string& candidate,
96 const PhoneNumberUtil& util) {
97 // The characters 'x' and 'X' can be (1) a carrier code, in which case they
98 // always precede the national significant number or (2) an extension sign,
99 // in which case they always precede the extension number. We assume a
100 // carrier code is more than 1 digit, so the first case has to have more than
101 // 1 consecutive 'x' or 'X', whereas the second case can only have exactly 1
104 found = candidate.find_first_of("xX");
105 // We ignore the character if 'x' or 'X' appears as the last character of
107 while (found != string::npos && found < candidate.length() - 1) {
108 // We only look for 'x' or 'X' in ASCII form.
109 char next_char = candidate[found + 1];
110 if (next_char == 'x' || next_char == 'X') {
111 // This is the carrier code case, in which the 'X's always precede the
112 // national significant number.
114 if (util.IsNumberMatchWithOneString(
115 number, candidate.substr(found, candidate.length() - found))
116 != PhoneNumberUtil::NSN_MATCH) {
120 string normalized_extension(candidate.substr(found,
121 candidate.length() - found));
122 util.NormalizeDigitsOnly(&normalized_extension);
123 if (normalized_extension != number.extension()) {
127 found = candidate.find_first_of("xX", found + 1);
133 #ifdef USE_GOOGLE_BASE
134 class PhoneNumberMatcherRegExps {
135 friend struct DefaultSingletonTraits<PhoneNumberMatcherRegExps>;
137 class PhoneNumberMatcherRegExps : public Singleton<PhoneNumberMatcherRegExps> {
138 friend class Singleton<PhoneNumberMatcherRegExps>;
139 #endif // USE_GOOGLE_BASE
141 string opening_parens_;
142 string closing_parens_;
144 // Limit on the number of pairs of brackets in a phone number.
145 string bracket_pair_limit_;
146 // Helper strings for the matching_brackets_ pattern.
147 // An opening bracket at the beginning may not be closed, but subsequent ones
148 // should be. It's also possible that the leading bracket was dropped, so we
149 // shouldn't be surprised if we see a closing bracket first.
150 string leading_maybe_matched_bracket_;
151 string bracket_pairs_;
152 // Limit on the number of leading (plus) characters.
154 // Limit on the number of consecutive punctuation characters.
155 string punctuation_limit_;
156 // The maximum number of digits allowed in a digit-separated block. As we
157 // allow all digits in a single block, this should be set high enough to
158 // accommodate the entire national number and the international country code.
159 int digit_block_limit_;
160 // Limit on the number of blocks separated by punctuation. Uses
161 // kDigitBlockLimit since some formats use spaces to separate each digit.
163 // A punctuation sequence allowing white space.
165 // A digits block without punctuation.
166 string digit_sequence_;
167 // Punctuation that may be at the start of a phone number - brackets and plus
169 string lead_class_chars_;
170 // Same as lead_class_chars_, but enclosed as a character class.
172 // Extra helper strings that form part of pattern_. These are stored
173 // separately since StrCat has a limit of 12 args.
174 string opening_punctuation_;
175 string optional_extn_pattern_;
178 // We use two different reg-ex factories here for performance reasons. RE2 is
179 // much faster for smaller reg-ex patterns, but the main pattern cannot be
180 // handled by RE2 in an efficient way.
181 scoped_ptr<const AbstractRegExpFactory> regexp_factory_for_pattern_;
182 scoped_ptr<const AbstractRegExpFactory> regexp_factory_;
184 // Matches strings that look like publication pages. Example:
185 // Computing Complete Answers to Queries in the Presence of Limited Access
186 // Patterns. Chen Li. VLDB J. 12(3): 211-227 (2003).
188 // The string "211-227 (2003)" is not a telephone number.
189 scoped_ptr<const RegExp> pub_pages_;
190 // Matches strings that look like dates using "/" as a separator. Examples:
191 // 3/10/2011, 31/10/96 or 08/31/95.
192 scoped_ptr<const RegExp> slash_separated_dates_;
193 // Matches timestamps. Examples: "2012-01-02 08:00". Note that the reg-ex does
194 // not include trailing ":\d\d" -- that is covered by time_stamps_suffix_.
195 scoped_ptr<const RegExp> time_stamps_;
196 scoped_ptr<const RegExp> time_stamps_suffix_;
197 // Pattern to check that brackets match. Opening brackets should be closed
198 // within a phone number. This also checks that there is something inside the
199 // brackets. Having no brackets at all is also fine.
200 scoped_ptr<const RegExp> matching_brackets_;
201 // Matches white-space, which may indicate the end of a phone number and the
202 // start of something else (such as a neighbouring zip-code). If white-space
203 // is found, continues to match all characters that are not typically used to
204 // start a phone number.
205 scoped_ptr<const RegExp> group_separator_;
206 scoped_ptr<const RegExp> capture_up_to_second_number_start_pattern_;
207 scoped_ptr<const RegExp> capturing_ascii_digits_pattern_;
208 // Compiled reg-ex representing lead_class_;
209 scoped_ptr<const RegExp> lead_class_pattern_;
210 // Phone number pattern allowing optional punctuation.
211 scoped_ptr<const RegExp> pattern_;
213 #ifdef USE_GOOGLE_BASE
214 static PhoneNumberMatcherRegExps* GetInstance() {
215 return Singleton<PhoneNumberMatcherRegExps>::get();
217 #endif // USE_GOOGLE_BASE
219 PhoneNumberMatcherRegExps()
220 : opening_parens_("(\\[\xEF\xBC\x88\xEF\xBC\xBB" /* "(\\[([" */),
221 closing_parens_(")\\]\xEF\xBC\x89\xEF\xBC\xBD" /* ")\\])]" */),
222 non_parens_(StrCat("[^", opening_parens_, closing_parens_, "]")),
223 bracket_pair_limit_(Limit(0, 3)),
224 leading_maybe_matched_bracket_(StrCat(
225 "(?:[", opening_parens_, "])?",
226 "(?:", non_parens_, "+[", closing_parens_, "])?")),
227 bracket_pairs_(StrCat(
228 "(?:[", opening_parens_, "]", non_parens_, "+",
229 "[", closing_parens_, "])", bracket_pair_limit_)),
230 lead_limit_(Limit(0, 2)),
231 punctuation_limit_(Limit(0, 4)),
232 digit_block_limit_(PhoneNumberUtil::kMaxLengthForNsn +
233 PhoneNumberUtil::kMaxLengthCountryCode),
234 block_limit_(Limit(0, digit_block_limit_)),
235 punctuation_(StrCat("[", PhoneNumberUtil::kValidPunctuation, "]",
236 punctuation_limit_)),
237 digit_sequence_(StrCat("\\p{Nd}", Limit(1, digit_block_limit_))),
238 lead_class_chars_(StrCat(opening_parens_, PhoneNumberUtil::kPlusChars)),
239 lead_class_(StrCat("[", lead_class_chars_, "]")),
240 opening_punctuation_(StrCat("(?:", lead_class_, punctuation_, ")")),
241 optional_extn_pattern_(StrCat(
243 PhoneNumberUtil::GetInstance()->GetExtnPatternsForMatching(),
245 regexp_factory_for_pattern_(new ICURegExpFactory()),
247 regexp_factory_(new RE2RegExpFactory()),
249 regexp_factory_(new ICURegExpFactory()),
251 pub_pages_(regexp_factory_->CreateRegExp(
252 "\\d{1,5}-+\\d{1,5}\\s{0,4}\\(\\d{1,4}")),
253 slash_separated_dates_(regexp_factory_->CreateRegExp(
254 "(?:(?:[0-3]?\\d/[01]?\\d)|"
255 "(?:[01]?\\d/[0-3]?\\d))/(?:[12]\\d)?\\d{2}")),
256 time_stamps_(regexp_factory_->CreateRegExp(
257 "[12]\\d{3}[-/]?[01]\\d[-/]?[0-3]\\d [0-2]\\d$")),
258 time_stamps_suffix_(regexp_factory_->CreateRegExp(":[0-5]\\d")),
259 matching_brackets_(regexp_factory_->CreateRegExp(
260 StrCat(leading_maybe_matched_bracket_, non_parens_, "+",
261 bracket_pairs_, non_parens_, "*"))),
262 group_separator_(regexp_factory_->CreateRegExp(
263 StrCat("\\p{Z}", "[^", lead_class_chars_, "\\p{Nd}]*"))),
264 capture_up_to_second_number_start_pattern_(
265 regexp_factory_->CreateRegExp(
266 PhoneNumberUtil::kCaptureUpToSecondNumberStart)),
267 capturing_ascii_digits_pattern_(
268 regexp_factory_->CreateRegExp("(\\d+)")),
269 lead_class_pattern_(regexp_factory_->CreateRegExp(lead_class_)),
270 pattern_(regexp_factory_for_pattern_->CreateRegExp(
271 StrCat("(", opening_punctuation_, lead_limit_,
272 digit_sequence_, "(?:", punctuation_, digit_sequence_, ")",
273 block_limit_, optional_extn_pattern_, ")"))) {
277 DISALLOW_COPY_AND_ASSIGN(PhoneNumberMatcherRegExps);
280 PhoneNumberMatcher::PhoneNumberMatcher(const PhoneNumberUtil& util,
282 const string& region_code,
283 PhoneNumberMatcher::Leniency leniency,
285 : reg_exps_(PhoneNumberMatcherRegExps::GetInstance()),
288 preferred_region_(region_code),
290 max_tries_(max_tries),
296 PhoneNumberMatcher::PhoneNumberMatcher(const string& text,
297 const string& region_code)
298 : reg_exps_(PhoneNumberMatcherRegExps::GetInstance()),
299 phone_util_(*PhoneNumberUtil::GetInstance()),
301 preferred_region_(region_code),
303 max_tries_(numeric_limits<int>::max()),
309 PhoneNumberMatcher::~PhoneNumberMatcher() {
313 bool PhoneNumberMatcher::IsLatinLetter(char32 letter) {
314 // Combining marks are a subset of non-spacing-mark.
315 if (!u_isalpha(letter) && (u_charType(letter) != U_NON_SPACING_MARK)) {
318 UBlockCode block = ublock_getCode(letter);
319 return ((block == UBLOCK_BASIC_LATIN) ||
320 (block == UBLOCK_LATIN_1_SUPPLEMENT) ||
321 (block == UBLOCK_LATIN_EXTENDED_A) ||
322 (block == UBLOCK_LATIN_EXTENDED_ADDITIONAL) ||
323 (block == UBLOCK_LATIN_EXTENDED_B) ||
324 (block == UBLOCK_COMBINING_DIACRITICAL_MARKS));
327 bool PhoneNumberMatcher::ParseAndVerify(const string& candidate, int offset,
328 PhoneNumberMatch* match) {
330 // Check the candidate doesn't contain any formatting which would indicate
331 // that it really isn't a phone number.
332 if (!reg_exps_->matching_brackets_->FullMatch(candidate)) {
336 // If leniency is set to VALID or stricter, we also want to skip numbers that
337 // are surrounded by Latin alphabetic characters, to skip cases like
338 // abc8005001234 or 8005001234def.
339 if (leniency_ >= VALID) {
340 // If the candidate is not at the start of the text, and does not start with
341 // phone-number punctuation, check the previous character.
342 scoped_ptr<RegExpInput> candidate_input(
343 reg_exps_->regexp_factory_->CreateInput(candidate));
345 !reg_exps_->lead_class_pattern_->Consume(candidate_input.get())) {
346 char32 previous_char;
347 const char* previous_char_ptr =
348 EncodingUtils::BackUpOneUTF8Character(text_.c_str(),
349 text_.c_str() + offset);
350 EncodingUtils::DecodeUTF8Char(previous_char_ptr, &previous_char);
351 // We return false if it is a latin letter or an invalid punctuation
353 if (IsInvalidPunctuationSymbol(previous_char) ||
354 IsLatinLetter(previous_char)) {
358 size_t lastCharIndex = offset + candidate.length();
359 if (lastCharIndex < text_.length()) {
361 const char* next_char_ptr =
362 EncodingUtils::AdvanceOneUTF8Character(
363 text_.c_str() + lastCharIndex - 1);
364 EncodingUtils::DecodeUTF8Char(next_char_ptr, &next_char);
365 if (IsInvalidPunctuationSymbol(next_char) || IsLatinLetter(next_char)) {
372 if (phone_util_.ParseAndKeepRawInput(candidate, preferred_region_, &number) !=
373 PhoneNumberUtil::NO_PARSING_ERROR) {
376 if (VerifyAccordingToLeniency(leniency_, number, candidate)) {
377 match->set_start(offset);
378 match->set_raw_string(candidate);
379 // We used ParseAndKeepRawInput to create this number, but for now we don't
380 // return the extra values parsed. TODO: stop clearing all values here and
381 // switch all users over to using raw_input() rather than the raw_string()
382 // of PhoneNumberMatch.
383 number.clear_country_code_source();
384 number.clear_preferred_domestic_carrier_code();
385 number.clear_raw_input();
386 match->set_number(number);
392 // Helper method to replace the verification method for each enum in the Java
394 bool PhoneNumberMatcher::VerifyAccordingToLeniency(
395 Leniency leniency, const PhoneNumber& number,
396 const string& candidate) const {
398 case PhoneNumberMatcher::POSSIBLE:
399 return phone_util_.IsPossibleNumber(number);
400 case PhoneNumberMatcher::VALID:
401 if (!phone_util_.IsValidNumber(number) ||
402 !ContainsOnlyValidXChars(number, candidate, phone_util_)) {
405 return IsNationalPrefixPresentIfRequired(number);
406 case PhoneNumberMatcher::STRICT_GROUPING: {
407 if (!phone_util_.IsValidNumber(number) ||
408 !ContainsOnlyValidXChars(number, candidate, phone_util_) ||
409 // Two or more slashes were present.
410 (FindNth(candidate, '/', 2) != string::npos) ||
411 !IsNationalPrefixPresentIfRequired(number)) {
414 // TODO(lararennie,shaopengjia): Evaluate how this works for other locales
415 // (testing has been limited to NANPA regions) and optimise if necessary.
416 string normalized_candidate =
417 NormalizeUTF8::NormalizeDecimalDigits(candidate);
418 vector<string> formatted_number_groups;
419 GetNationalNumberGroups(phone_util_, number, &formatted_number_groups);
420 size_t from_index = 0;
421 // Check each group of consecutive digits are not broken into separate
422 // groups in the normalized_candidate string.
423 for (size_t i = 0; i < formatted_number_groups.size(); ++i) {
424 // Fails if the substring of normalized_candidate starting from
425 // from_index doesn't contain the consecutive digits in digit_group.
426 from_index = normalized_candidate.find(formatted_number_groups.at(i),
428 if (from_index == string::npos) {
431 // Moves from_index forward.
432 from_index += formatted_number_groups.at(i).length();
433 if (i == 0 && from_index < normalized_candidate.length()) {
434 // We are at the position right after the NDC. Note although
435 // normalized_candidate might contain non-ASCII formatting characters,
436 // they won't be treated as ASCII digits when converted to a char.
437 if (isdigit(normalized_candidate.at(from_index))) {
438 // This means there is no formatting symbol after the NDC. In this
439 // case, we only accept the number if there is no formatting
440 // symbol at all in the number, except for extensions.
441 string national_significant_number;
442 phone_util_.GetNationalSignificantNumber(
443 number, &national_significant_number);
444 return HasPrefixString(
445 normalized_candidate.substr(
446 from_index - formatted_number_groups.at(i).length()),
447 national_significant_number);
451 // The check here makes sure that we haven't mistakenly already used the
452 // extension to match the last group of the subscriber number. Note the
453 // extension cannot have formatting in-between digits.
455 normalized_candidate.substr(from_index).find(number.extension()) !=
458 case PhoneNumberMatcher::EXACT_GROUPING: {
459 if (!phone_util_.IsValidNumber(number) ||
460 !ContainsOnlyValidXChars(number, candidate, phone_util_) ||
461 // Two or more slashes were present.
462 (FindNth(candidate, '/', 2) != string::npos) ||
463 !IsNationalPrefixPresentIfRequired(number)) {
466 // TODO(lararennie,shaopengjia): Evaluate how this works for other locales
467 // (testing has been limited to NANPA regions) and optimise if necessary.
468 vector<string> candidate_groups;
469 string normalized_candidate =
470 NormalizeUTF8::NormalizeDecimalDigits(candidate);
471 const scoped_ptr<RegExpInput> candidate_number(
472 reg_exps_->regexp_factory_->CreateInput(normalized_candidate));
474 while (reg_exps_->capturing_ascii_digits_pattern_->FindAndConsume(
475 candidate_number.get(),
477 candidate_groups.push_back(digit_block);
480 // Set this to the last group, skipping it if the number has an extension.
481 int candidate_number_group_index =
482 number.has_extension() ? candidate_groups.size() - 2
483 : candidate_groups.size() - 1;
484 // First we check if the national significant number is formatted as a
485 // block. We use contains and not equals, since the national significant
486 // number may be present with a prefix such as a national number prefix,
487 // or the country code itself.
488 string national_significant_number;
489 phone_util_.GetNationalSignificantNumber(number,
490 &national_significant_number);
491 if (candidate_groups.size() == 1 ||
492 candidate_groups.at(candidate_number_group_index).find(
493 national_significant_number) != string::npos) {
496 vector<string> formatted_number_groups;
497 GetNationalNumberGroups(phone_util_, number, &formatted_number_groups);
498 // Starting from the end, go through in reverse, excluding the first
499 // group, and check the candidate and number groups are the same.
500 for (int formatted_number_group_index =
501 (formatted_number_groups.size() - 1);
502 formatted_number_group_index > 0 &&
503 candidate_number_group_index >= 0;
504 --formatted_number_group_index, --candidate_number_group_index) {
505 if (candidate_groups.at(candidate_number_group_index) !=
506 formatted_number_groups.at(formatted_number_group_index)) {
510 // Now check the first group. There may be a national prefix at the start,
511 // so we only check that the candidate group ends with the formatted
513 return (candidate_number_group_index >= 0 &&
514 HasSuffixString(candidate_groups.at(candidate_number_group_index),
515 formatted_number_groups.at(0)));
518 LOG(ERROR) << "No implementation defined for verification for leniency "
519 << static_cast<int>(leniency);
524 bool PhoneNumberMatcher::ExtractInnerMatch(const string& candidate, int offset,
525 PhoneNumberMatch* match) {
527 // Try removing either the first or last "group" in the number and see if this
528 // gives a result. We consider white space to be a possible indication of
529 // the start or end of the phone number.
530 scoped_ptr<RegExpInput> candidate_input(
531 reg_exps_->regexp_factory_->CreateInput(candidate));
532 if (reg_exps_->group_separator_->FindAndConsume(candidate_input.get(),
534 // Try the first group by itself.
535 int group_start_index =
536 candidate.length() - candidate_input->ToString().length();
537 string first_group_only = candidate.substr(0, group_start_index);
538 phone_util_.TrimUnwantedEndChars(&first_group_only);
539 bool success = ParseAndVerify(first_group_only, offset, match);
545 // Try the rest of the candidate without the first group.
546 string without_first_group(candidate_input->ToString());
547 phone_util_.TrimUnwantedEndChars(&without_first_group);
549 ParseAndVerify(without_first_group, offset + group_start_index, match);
555 if (max_tries_ > 0) {
556 while (reg_exps_->group_separator_->FindAndConsume(candidate_input.get(),
558 // Find the last group.
560 int last_group_start =
561 candidate.length() - candidate_input->ToString().length();
562 string without_last_group = candidate.substr(0, last_group_start);
563 phone_util_.TrimUnwantedEndChars(&without_last_group);
564 if (without_last_group == first_group_only) {
565 // If there are only two groups, then the group "without the last group"
566 // is the same as the first group. In these cases, we don't want to
567 // re-check the number group, so we exit already.
570 success = ParseAndVerify(without_last_group, offset, match);
580 bool PhoneNumberMatcher::ExtractMatch(const string& candidate, int offset,
581 PhoneNumberMatch* match) {
583 // Skip a match that is more likely a publication page reference or a date.
584 if (reg_exps_->pub_pages_->PartialMatch(candidate) ||
585 reg_exps_->slash_separated_dates_->PartialMatch(candidate)) {
588 // Skip potential time-stamps.
589 if (reg_exps_->time_stamps_->PartialMatch(candidate)) {
590 scoped_ptr<RegExpInput> following_text(
591 reg_exps_->regexp_factory_->CreateInput(
592 text_.substr(offset + candidate.size())));
593 if (reg_exps_->time_stamps_suffix_->Consume(following_text.get())) {
598 // Try to come up with a valid match given the entire candidate.
599 if (ParseAndVerify(candidate, offset, match)) {
603 // If that failed, try to find an "inner match" - there might be a phone
604 // number within this candidate.
605 return ExtractInnerMatch(candidate, offset, match);
608 bool PhoneNumberMatcher::HasNext() {
609 if (state_ == NOT_READY) {
610 PhoneNumberMatch temp_match;
611 if (!Find(search_index_, &temp_match)) {
614 last_match_.reset(new PhoneNumberMatch(temp_match.start(),
615 temp_match.raw_string(),
616 temp_match.number()));
617 search_index_ = last_match_->end();
621 return state_ == READY;
624 bool PhoneNumberMatcher::Next(PhoneNumberMatch* match) {
626 // Check the state and find the next match as a side-effect if necessary.
630 match->CopyFrom(*last_match_);
632 last_match_.reset(NULL);
636 bool PhoneNumberMatcher::Find(int index, PhoneNumberMatch* match) {
639 scoped_ptr<RegExpInput> text(
640 reg_exps_->regexp_factory_for_pattern_->CreateInput(text_.substr(index)));
642 while ((max_tries_ > 0) &&
643 reg_exps_->pattern_->FindAndConsume(text.get(), &candidate)) {
644 int start = text_.length() - text->ToString().length() - candidate.length();
645 // Check for extra numbers at the end.
646 reg_exps_->capture_up_to_second_number_start_pattern_->
647 PartialMatch(candidate, &candidate);
648 if (ExtractMatch(candidate, start, match)) {
652 index = start + candidate.length();
658 bool PhoneNumberMatcher::IsNationalPrefixPresentIfRequired(
659 const PhoneNumber& number) const {
660 // First, check how we deduced the country code. If it was written in
661 // international format, then the national prefix is not required.
662 if (number.country_code_source() != PhoneNumber::FROM_DEFAULT_COUNTRY) {
665 string phone_number_region;
666 phone_util_.GetRegionCodeForCountryCode(
667 number.country_code(), &phone_number_region);
668 const PhoneMetadata* metadata =
669 phone_util_.GetMetadataForRegion(phone_number_region);
673 // Check if a national prefix should be present when formatting this number.
674 string national_number;
675 phone_util_.GetNationalSignificantNumber(number, &national_number);
676 const NumberFormat* format_rule =
677 phone_util_.ChooseFormattingPatternForNumber(metadata->number_format(),
679 // To do this, we check that a national prefix formatting rule was present and
680 // that it wasn't just the first-group symbol ($1) with punctuation.
681 if (format_rule && !format_rule->national_prefix_formatting_rule().empty()) {
682 if (format_rule->national_prefix_optional_when_formatting()) {
683 // The national-prefix is optional in these cases, so we don't need to
684 // check if it was present.
687 // Remove the first-group symbol.
688 string candidate_national_prefix_rule(
689 format_rule->national_prefix_formatting_rule());
690 // We assume that the first-group symbol will never be _before_ the national
692 candidate_national_prefix_rule.erase(
693 candidate_national_prefix_rule.find("$1"));
694 phone_util_.NormalizeDigitsOnly(&candidate_national_prefix_rule);
695 if (candidate_national_prefix_rule.empty()) {
696 // National Prefix not needed for this number.
699 // Normalize the remainder.
700 string raw_input_copy(number.raw_input());
701 // Check if we found a national prefix and/or carrier code at the start of
702 // the raw input, and return the result.
703 phone_util_.NormalizeDigitsOnly(&raw_input_copy);
704 return phone_util_.MaybeStripNationalPrefixAndCarrierCode(
707 NULL); // Don't need to keep the stripped carrier code.
712 } // namespace phonenumbers