@recaptime-dev's working patches + fork for Phorge, a community fork of Phabricator. (Upstream dev and stable branches are at upstream/main and upstream/stable respectively.)
hq.recaptime.dev/wiki/Phorge
phorge
phabricator
Implement the "present" and "absent" operators in the Ferret execution engine
Summary:
Ref T13509. Now that the compiler can parse these queries, actually implement them.
These are fairly easy to implement:
- For present, just "JOIN". If it works, the field is present.
- For absent, we "LEFT JOIN" and then "WHERE any_column IS NULL".
Test Plan: Searched for various documents with and without fields present, got sensible results in Maniphest. For example, "body:-" finds tasks with no body, "body:- duck" finds tasks with no body and "duck" elsewhere in the content, and so on.
Maniphest Tasks: T13509
Differential Revision: https://secure.phabricator.com/D21110
+52
-8
+52
-8
src/infrastructure/query/policy/PhabricatorCursorPagedPolicyAwareQuery.php
+52
-8
src/infrastructure/query/policy/PhabricatorCursorPagedPolicyAwareQuery.php
···
1801
1801
$this->ferretEngine = $engine;
1802
1802
$this->ferretTokens = $fulltext_tokens;
1803
1803
1804
+
$op_absent = PhutilSearchQueryCompiler::OPERATOR_ABSENT;
1805
+
1804
1806
$default_function = $engine->getDefaultFunctionKey();
1805
1807
$table_map = array();
1806
1808
$idx = 1;
1807
1809
foreach ($this->ferretTokens as $fulltext_token) {
1808
1810
$raw_token = $fulltext_token->getToken();
1809
-
$function = $raw_token->getFunction();
1810
1811
1812
+
$function = $raw_token->getFunction();
1811
1813
if ($function === null) {
1812
1814
$function = $default_function;
1813
1815
}
1814
1816
1815
1817
$raw_field = $engine->getFieldForFunction($function);
1816
1818
1819
+
// NOTE: The query compiler guarantees that a query can not make a
1820
+
// field both "present" and "absent", so it's safe to just use the
1821
+
// first operator we encounter to determine whether the table is
1822
+
// optional or not.
1823
+
1824
+
$operator = $raw_token->getOperator();
1825
+
$is_optional = ($operator === $op_absent);
1826
+
1817
1827
if (!isset($table_map[$function])) {
1818
1828
$alias = 'ftfield_'.$idx++;
1819
1829
$table_map[$function] = array(
1820
1830
'alias' => $alias,
1821
1831
'key' => $raw_field,
1832
+
'optional' => $is_optional,
1822
1833
);
1823
1834
}
1824
1835
}
···
1966
1977
1967
1978
$op_sub = PhutilSearchQueryCompiler::OPERATOR_SUBSTRING;
1968
1979
$op_not = PhutilSearchQueryCompiler::OPERATOR_NOT;
1980
+
$op_absent = PhutilSearchQueryCompiler::OPERATOR_ABSENT;
1981
+
$op_present = PhutilSearchQueryCompiler::OPERATOR_PRESENT;
1969
1982
1970
1983
$engine = $this->ferretEngine;
1971
1984
$stemmer = $engine->newStemmer();
···
1976
1989
foreach ($this->ferretTokens as $fulltext_token) {
1977
1990
$raw_token = $fulltext_token->getToken();
1978
1991
1992
+
$operator = $raw_token->getOperator();
1993
+
1979
1994
// If this is a negated term like "-pomegranate", don't join the ngram
1980
1995
// table since we aren't looking for documents with this term. (We could
1981
1996
// LEFT JOIN the table and require a NULL row, but this is probably more
1982
1997
// trouble than it's worth.)
1983
-
if ($raw_token->getOperator() == $op_not) {
1998
+
if ($operator === $op_not) {
1999
+
continue;
2000
+
}
2001
+
2002
+
// Neither the "present" or "absent" operators benefit from joining
2003
+
// the ngram table.
2004
+
if ($operator === $op_absent || $operator === $op_present) {
1984
2005
continue;
1985
2006
}
1986
2007
···
2143
2164
$op_sub = PhutilSearchQueryCompiler::OPERATOR_SUBSTRING;
2144
2165
$op_not = PhutilSearchQueryCompiler::OPERATOR_NOT;
2145
2166
$op_exact = PhutilSearchQueryCompiler::OPERATOR_EXACT;
2167
+
$op_absent = PhutilSearchQueryCompiler::OPERATOR_ABSENT;
2168
+
$op_present = PhutilSearchQueryCompiler::OPERATOR_PRESENT;
2146
2169
2147
2170
$where = array();
2148
-
$current_function = 'all';
2171
+
$default_function = $engine->getDefaultFunctionKey();
2149
2172
foreach ($this->ferretTokens as $fulltext_token) {
2150
2173
$raw_token = $fulltext_token->getToken();
2151
2174
$value = $raw_token->getValue();
2152
2175
2153
2176
$function = $raw_token->getFunction();
2154
2177
if ($function === null) {
2155
-
$function = $current_function;
2178
+
$function = $default_function;
2156
2179
}
2157
-
$current_function = $function;
2180
+
2181
+
$operator = $raw_token->getOperator();
2158
2182
2159
2183
$table_alias = $table_map[$function]['alias'];
2160
2184
2161
-
$is_not = ($raw_token->getOperator() == $op_not);
2185
+
// If this is a "field is present" operator, we've already implicitly
2186
+
// guaranteed this by JOINing the table. We don't need to do any
2187
+
// more work.
2188
+
$is_present = ($operator === $op_present);
2189
+
if ($is_present) {
2190
+
continue;
2191
+
}
2162
2192
2163
-
if ($raw_token->getOperator() == $op_sub) {
2193
+
// If this is a "field is absent" operator, we just want documents
2194
+
// which failed to match to a row when we LEFT JOINed the table. This
2195
+
// means there's no index for the field.
2196
+
$is_absent = ($operator === $op_absent);
2197
+
if ($is_absent) {
2198
+
$where[] = qsprintf(
2199
+
$conn,
2200
+
'(%T.rawCorpus IS NULL)',
2201
+
$table_alias);
2202
+
continue;
2203
+
}
2204
+
2205
+
$is_not = ($operator === $op_not);
2206
+
2207
+
if ($operator == $op_sub) {
2164
2208
$is_substring = true;
2165
2209
} else {
2166
2210
$is_substring = false;
2167
2211
}
2168
2212
2169
2213
// If we're doing exact search, just test the raw corpus.
2170
-
$is_exact = ($raw_token->getOperator() == $op_exact);
2214
+
$is_exact = ($operator === $op_exact);
2171
2215
if ($is_exact) {
2172
2216
if ($is_not) {
2173
2217
$where[] = qsprintf(