From b3310ec4e29ca2ef45e36a7b188503cd2637c8ba Mon Sep 17 00:00:00 2001 From: Matt Perry Date: Wed, 17 Dec 2008 00:50:50 +0000 Subject: [PATCH 13/16] [fts2] Interpret "foo*" as a prefix search. By default it interpreted it as "foo *" (two tokens). Original review URL: http://codereview.chromium.org/14176 --- third_party/sqlite/src/ext/fts2/fts2.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) diff --git a/third_party/sqlite/src/ext/fts2/fts2.c b/third_party/sqlite/src/ext/fts2/fts2.c index 944f324..1c68af5 100644 --- a/third_party/sqlite/src/ext/fts2/fts2.c +++ b/third_party/sqlite/src/ext/fts2/fts2.c @@ -3558,6 +3558,7 @@ static int tokenizeSegment( int firstIndex = pQuery->nTerms; int iCol; int nTerm = 1; + int iEndLast = -1; int rc = pModule->xOpen(pTokenizer, pSegment, nSegment, &pCursor); if( rc!=SQLITE_OK ) return rc; @@ -3582,6 +3583,20 @@ static int tokenizeSegment( pQuery->nextIsOr = 1; continue; } + + /* + * The ICU tokenizer considers '*' a break character, so the code below + * sets isPrefix correctly, but since that code doesn't eat the '*', the + * ICU tokenizer returns it as the next token. So eat it here until a + * better solution presents itself. + */ + if( pQuery->nTerms>0 && nToken==1 && pSegment[iBegin]=='*' && + iEndLast==iBegin){ + pQuery->pTerms[pQuery->nTerms-1].isPrefix = 1; + continue; + } + iEndLast = iEnd; + queryAdd(pQuery, pToken, nToken); if( !inPhrase && iBegin>0 && pSegment[iBegin-1]=='-' ){ pQuery->pTerms[pQuery->nTerms-1].isNot = 1; -- 2.2.1