fix address detection when the city and state are both valid states
This fixes detecting "4 E. 86th St New York, NY"
The old logic found 'New York' and assumed that St was the city
name. Since 86th is not a valid street suffix, the test failed.
The new logic looks for a valid street suffix instead of skipping
'city'. If it finds one, and the subsequent street suffix test
fails, it resets to search again for a valid state name while
retaining the starting point for the address (the street number).
Fixing this exposed a bug in the zip code detection code
where it dropped the first character of the actual state name.
diff --git a/WebKit/android/nav/CacheBuilder.cpp b/WebKit/android/nav/CacheBuilder.cpp
index 56cb2ae..037a6e5 100644
--- a/WebKit/android/nav/CacheBuilder.cpp
+++ b/WebKit/android/nav/CacheBuilder.cpp
@@ -1590,6 +1590,7 @@
s->mCurrentStart = chars;
s->mEnd = chars + length;
int candIndex = 0;
+ bool retryState;
bool mustBeAllUpper = false;
bool secondHalf = false;
chars -= 1;
@@ -1914,6 +1915,15 @@
s->mZipDelimiter = true;
} else if (WTF::isASCIIAlpha(ch) == false)
s->mZipDelimiter = true;
+ else {
+ if (s->mLetterCount == 0) {
+ s->mBases[s->mWordCount] = baseChars;
+ s->mWords[s->mWordCount] = chars;
+ s->mStarts[s->mWordCount] = s->mCurrentStart;
+ s->mUnparsed = true;
+ }
+ ++s->mLetterCount;
+ }
if (s->mNumberCount == 5 || s->mNumberCount == 9) {
if (validZip(s->mZipHint, s->mZipStart) == false)
goto noZipMatch;
@@ -1945,8 +1955,9 @@
break;
s->mProgress = FIND_STREET;
case FIND_STREET:
- findStreet: // minus two below skips city before state
- for (int wordsIndex = s->mStateWord - 2; wordsIndex >= 0; --wordsIndex) {
+ findStreet:
+ retryState = false;
+ for (int wordsIndex = s->mStateWord - 1; wordsIndex >= 0; --wordsIndex) {
const UChar* test = s->mWords[wordsIndex];
UChar letter = test[0];
letter -= 'A';
@@ -1992,17 +2003,27 @@
wordReduction = shift;
} while (s->mNumberWords >> ++shift != 0);
if (wordReduction >= 0) {
- if (s->mContinuationNode)
+ if (s->mContinuationNode) {
+ if (retryState)
+ break;
return FOUND_NONE;
+ }
s->mStartResult = s->mWords[wordReduction] - s->mStarts[wordReduction];
}
}
- return FOUND_COMPLETE;
+ if (wordsIndex != s->mStateWord - 1)
+ return FOUND_COMPLETE;
+ retryState = true;
}
nextTest:
names += offset;
}
}
+ if (retryState) {
+ s->mProgress = ADDRESS_LINE;
+ s->mStates = NULL;
+ continue;
+ }
if (s->mNumberWords != 0) {
unsigned shift = 0;
while ((s->mNumberWords & (1 << shift)) == 0)
