Disallow non-utf8 chars as input to phonenumbermatcher API (#2707)

cpp/src/phonenumbers/phonenumbermatcher.cc

@@ -33,7 +33,6 @@
 #include <string>
 #include <utility>
 #include <vector>
-
 #include <unicode/uchar.h>
 
 #include "phonenumbers/alternate_format.h"
@@ -52,6 +51,7 @@
 #include "phonenumbers/regexp_adapter_icu.h"
 #include "phonenumbers/regexp_cache.h"
 #include "phonenumbers/stringutil.h"
+#include "phonenumbers/utf/unicodetext.h"
 
 #ifdef I18N_PHONENUMBERS_USE_RE2
 #include "phonenumbers/regexp_adapter_re2.h"
@@ -407,7 +407,9 @@ PhoneNumberMatcher::PhoneNumberMatcher(const PhoneNumberUtil& util,
       max_tries_(max_tries),
       state_(NOT_READY),
       last_match_(NULL),
-      search_index_(0) {
+      search_index_(0),
+      is_input_valid_utf8_(true) {
+  is_input_valid_utf8_ = IsInputUtf8(); 
 }
 
 PhoneNumberMatcher::PhoneNumberMatcher(const string& text,
@@ -421,12 +423,20 @@ PhoneNumberMatcher::PhoneNumberMatcher(const string& text,
       max_tries_(numeric_limits<int>::max()),
       state_(NOT_READY),
       last_match_(NULL),
-      search_index_(0) {
+      search_index_(0),
+      is_input_valid_utf8_(true) {
+  is_input_valid_utf8_ =  IsInputUtf8();
 }
 
 PhoneNumberMatcher::~PhoneNumberMatcher() {
 }
 
+bool PhoneNumberMatcher::IsInputUtf8() {
+  UnicodeText number_as_unicode;
+  number_as_unicode.PointToUTF8(text_.c_str(), text_.size());
+  return number_as_unicode.UTF8WasValid();
+}
+
 // static
 bool PhoneNumberMatcher::IsLatinLetter(char32 letter) {
   // Combining marks are a subset of non-spacing-mark.
@@ -626,6 +636,11 @@ bool PhoneNumberMatcher::ExtractMatch(const string& candidate, int offset,
 }
 
 bool PhoneNumberMatcher::HasNext() {
+  // Input should contain only UTF-8 characters.
+  if (!is_input_valid_utf8_) {
+    state_ = DONE;
+    return false;
+  }
   if (state_ == NOT_READY) {
     PhoneNumberMatch temp_match;
     if (!Find(search_index_, &temp_match)) {

cpp/src/phonenumbers/phonenumbermatcher.h

@@ -87,7 +87,8 @@ class PhoneNumberMatcher {
 
   ~PhoneNumberMatcher();
 
-  // Returns true if the text sequence has another match.
+  // Returns true if the text sequence has another match. Return false if not.
+  // Always returns false when input contains non UTF-8 characters.
   bool HasNext();
 
   // Gets next match from text sequence.
@@ -101,6 +102,9 @@ class PhoneNumberMatcher {
     DONE,
   };
 
+  // Checks if the to check if the provided text_ is in UTF-8 or not.
+  bool IsInputUtf8();
+
   // Attempts to extract a match from a candidate string. Returns true if a
   // match is found, otherwise returns false. The value "offset" refers to the
   // start index of the candidate string within the overall text.
@@ -202,6 +206,9 @@ class PhoneNumberMatcher {
   // The next index to start searching at. Undefined in State.DONE.
   int search_index_;
 
+  // Flag to set or check if input text is in UTF-8 or not.
+  bool is_input_valid_utf8_;
+
   DISALLOW_COPY_AND_ASSIGN(PhoneNumberMatcher);
 };
 

cpp/test/phonenumbers/phonenumbermatcher_test.cc

@@ -1078,6 +1078,23 @@ TEST_F(PhoneNumberMatcherTest, NoMatchIfNoNumber) {
   EXPECT_FALSE(matcher->HasNext());
 }
 
+TEST_F(PhoneNumberMatcherTest, NoErrorWithSpecialCharacters) {
+  string stringWithSpecialCharacters =
+      "Myfuzzvar1152: \"My info:%415-666-7777 123 fake street\"\nfuzzvar1155: "
+      "47\nfuzzvar1158: %415-666-1234 "
+      "i18n_phonenumbers_Pho\356eNumberMatcher_Leniency_VALID_1"
+      "\nfuzzvar1159: 20316 info:%415-666-7777 123 fake str79ee\nt";
+  string Numbers;
+  for (int i = 0; i < 100; ++i)
+    Numbers.append(stringWithSpecialCharacters);
+  scoped_ptr<PhoneNumberMatcher> matcher(
+      GetMatcherWithLeniency(Numbers, RegionCode::US(),
+                             PhoneNumberMatcher::POSSIBLE));
+  // Since the input text contains invalid UTF-8, we do not return
+  // any matches.
+  EXPECT_FALSE(matcher->HasNext());
+}
+
 TEST_F(PhoneNumberMatcherTest, Sequences) {
   // Test multiple occurrences.
   const string text = "Call 033316005  or 032316005!";