Explicit "~" does not look for synonym when using FLAG_PARTIAL

[olinux]

Query parser does not look for synonyms when FLAG_PARTIAL is combined with FLAG_AUTO_SYNONYMS or FLAG_SYNONYM

Example queries

• hello
• ~hello
• hello world
• ~hello world
• world ~hello

Example

Create database, add synonyms and parse above queries

Problems are indicated by "PROBLEM" in comments below.

// create database
$xapianDatabasePath = '/tmp/xapianSynonymTest';
$db = new XapianWritableDatabase($xapianDatabasePath, Xapian::DB_CREATE_OR_OVERWRITE);

$db->add_synonym('hello', 'hi');
$db->add_synonym('hello', 'howdy');

$db = null;

// run some queries
$db = new XapianDatabase($xapianDatabasePath);
$enquire = new XapianEnquire($db);
$qp = new XapianQueryParser();
$stemmer = new XapianStem('english');
$qp->set_stemmer($stemmer);
$qp->set_database($db);

$xapianFlags = XapianQueryParser::FLAG_BOOLEAN | XapianQueryParser::FLAG_AUTO_SYNONYMS;
echo $xapianFlags . '<br>';

// works as expected
// Xapian::Query((hello:(pos=1) OR hi:(pos=1) OR howdy:(pos=1)))
$query = $qp->parse_query('hello', $xapianFlags);
echo $query->get_description() . '<br>';

// works as expected
// Xapian::Query((hello:(pos=1) OR hi:(pos=1) OR howdy:(pos=1)))
$query = $qp->parse_query('~hello', $xapianFlags);
echo $query->get_description() . '<br>';

// works as expected
// Xapian::Query((hello:(pos=1) OR hi:(pos=1) OR howdy:(pos=1) OR world:(pos=2)))
$query = $qp->parse_query('hello world', $xapianFlags);
echo $query->get_description() . '<br>';

// works as expected
// Xapian::Query((hello:(pos=1) OR hi:(pos=1) OR howdy:(pos=1) OR world:(pos=2)))
$query = $qp->parse_query('~hello world', $xapianFlags);
echo $query->get_description() . '<br>';

$xapianFlags = XapianQueryParser::FLAG_BOOLEAN | XapianQueryParser::FLAG_PARTIAL | XapianQueryParser::FLAG_AUTO_SYNONYMS;
echo $xapianFlags . '<br>';

// works as expected
// Xapian::Query(hello:(pos=1))
$query = $qp->parse_query('hello', $xapianFlags);
echo $query->get_description() . '<br>';

// PROBLEM - EXPLICIT "~" SHOULD FORCE LOOK FOR A SYNONYM
// Xapian::Query(hello:(pos=1))
$query = $qp->parse_query('~hello', $xapianFlags);
echo $query->get_description() . '<br>';

// works as expected
// Xapian::Query((hello:(pos=1) OR hi:(pos=1) OR howdy:(pos=1) OR world:(pos=2)))
$query = $qp->parse_query('hello world', $xapianFlags);
echo $query->get_description() . '<br>';

// works as expected
// Xapian::Query((hello:(pos=1) OR hi:(pos=1) OR howdy:(pos=1) OR world:(pos=2)))
$query = $qp->parse_query('~hello world', $xapianFlags);
echo $query->get_description() . '<br>';

// PROBLEM - EXPLICIT "~" SHOULD FORCE LOOK FOR A SYNONYM
// Xapian::Query((hello:(pos=1) OR hi:(pos=1) OR howdy:(pos=1) OR world:(pos=2)))
$query = $qp->parse_query('world ~hello', $xapianFlags);
echo $query->get_description() . '<br>';

$xapianFlags = XapianQueryParser::FLAG_BOOLEAN | XapianQueryParser::FLAG_PARTIAL | XapianQueryParser::FLAG_SYNONYM;
echo $xapianFlags . '<br>';

// works as expected
// Xapian::Query(hello:(pos=1))
$query = $qp->parse_query('hello', $xapianFlags);
echo $query->get_description() . '<br>';

// PROBLEM - EXPLICIT "~" SHOULD FORCE LOOK FOR A SYNONYM
// Xapian::Query(hello:(pos=1))
$query = $qp->parse_query('~hello', $xapianFlags);
echo $query->get_description() . '<br>';

// works as expected
// Xapian::Query((hello:(pos=1) OR world:(pos=2)))
$query = $qp->parse_query('hello world', $xapianFlags);
echo $query->get_description() . '<br>';

// works as expected
// Xapian::Query((hello:(pos=1) OR hi:(pos=1) OR howdy:(pos=1) OR world:(pos=2)))
$query = $qp->parse_query('~hello world', $xapianFlags);
echo $query->get_description() . '<br>';

[Olly Betts]

Under: FLAG_BOOLEAN | FLAG_PARTIAL | FLAG_AUTO_SYNONYMS

// PROBLEM - EXPLICIT "~" SHOULD FORCE LOOK FOR A SYNONYM
// Xapian::Query(hello:(pos=1))
$query = $qp->parse_query('~hello', $xapianFlags);

I'm not sure I agree here. You didn't pass FLAG_SYNONYM, so the "~" synonym operator isn't enabled, and the "~" is ignored as a spurious character (like "%" would be in "%hello", for example). So this case is handled just as 'hello' is.

If you want "~" to be recognised as meaning "synonym this explicitly", I think you should be passing FLAG_SYNONYM. FLAG_AUTO_SYNONYMS could imply FLAG_SYNONYM, but then we lose the flexibility of not having "~" recognised specially.

Similar arguments apply to: world ~hello

The other cases are fixed in trunk r13611.

Marking for backport to 1.1.17.

[Olly Betts]

Backported for 1.1.17 in r13613.