Provide some "term vs substring" support for the Ferret engine

Summary:
Ref T12819. Distinguishes between "term" queries and "substring" queries, and tries to match them correctly most of the time. For example:

  - `example` matches "example", obviously.
  - `~amp` matches "example", but `amp` does not.
  - `examples` matches "example" through stemming.
  - `"examples"` does not match "example" (quoted text does not stem).
  - `"an examp"` does not match "an example" (quoted text is still term text).
  - `~"an examp"` matches "an example" (quoted, substring-operator text uses substring search).

Test Plan: Ran searches similar to the above, they seemed to do what they should.

Reviewers: chad

Reviewed By: chad

Maniphest Tasks: T12819

Differential Revision: https://secure.phabricator.com/D18500

src/applications/maniphest/query/ManiphestTaskSearchEngine.php

@@ -92,8 +92,8 @@ final class ManiphestTaskSearchEngine
         ->setLabel(pht('Contains Words'))
         ->setKey('fulltext'),
       id(new PhabricatorSearchTextField())
-        ->setLabel(pht('Matches (Prototype)'))
-        ->setKey('ferret')
+        ->setLabel(pht('Query (Prototype)'))
+        ->setKey('query')
         ->setIsHidden($hide_ferret),
       id(new PhabricatorSearchThreeStateField())
         ->setLabel(pht('Open Parents'))
@@ -150,8 +150,8 @@ final class ManiphestTaskSearchEngine
       'statuses',
       'priorities',
       'subtypes',
+      'query',
       'fulltext',
-      'ferret',
       'hasParents',
       'hasSubtasks',
       'parentIDs',
@@ -231,8 +231,8 @@ final class ManiphestTaskSearchEngine
       $query->withFullTextSearch($map['fulltext']);
     }
 
-    if (strlen($map['ferret'])) {
-      $raw_query = $map['ferret'];
+    if (strlen($map['query'])) {
+      $raw_query = $map['query'];
 
       $compiler = id(new PhutilSearchQueryCompiler())
         ->setEnableFunctions(true);

src/applications/search/engineextension/PhabricatorFerretFulltextEngineExtension.php

@@ -50,9 +50,11 @@ final class PhabricatorFerretFulltextEngineExtension
         continue;
       }
 
-      $normal_corpus = $stemmer->stemCorpus($raw_corpus);
       $term_corpus = $ngram_engine->newTermsCorpus($raw_corpus);
 
+      $normal_corpus = $stemmer->stemCorpus($raw_corpus);
+      $normal_coprus = $ngram_engine->newTermsCorpus($normal_corpus);
+
       if (!isset($ferret_corpus_map[$key])) {
         $ferret_corpus_map[$key] = $empty_template;
       }
@@ -67,16 +69,23 @@ final class PhabricatorFerretFulltextEngineExtension
     }
 
     $ferret_fields = array();
+    $ngrams_source = array();
     foreach ($ferret_corpus_map as $key => $fields) {
       $raw_corpus = $fields['raw'];
       $raw_corpus = implode("\n", $raw_corpus);
+      $ngrams_source[] = $raw_corpus;
 
       $normal_corpus = $fields['normal'];
-      $normal_corpus = implode("\n", $normal_corpus);
+      $normal_corpus = implode(' ', $normal_corpus);
+      if (strlen($normal_corpus)) {
+        $ngrams_source[] = $normal_corpus;
+        $normal_corpus = ' '.$normal_corpus.' ';
+      }
 
       $term_corpus = $fields['term'];
       $term_corpus = implode(' ', $term_corpus);
       if (strlen($term_corpus)) {
+        $ngrams_source[] = $term_corpus;
         $term_corpus = ' '.$term_corpus.' ';
       }
 
@@ -86,9 +95,7 @@ final class PhabricatorFerretFulltextEngineExtension
         ->setTermCorpus($term_corpus)
         ->setNormalCorpus($normal_corpus);
     }
-
-    $ngrams_source = $ferret_corpus_map[$key_all]['raw'];
-    $ngrams_source = implode("\n", $ngrams_source);
+    $ngrams_source = implode(' ', $ngrams_source);
 
     $ngrams = $ngram_engine->getNgramsFromString($ngrams_source, 'index');
 

src/infrastructure/query/policy/PhabricatorCursorPagedPolicyAwareQuery.php

@@ -1409,8 +1409,11 @@ abstract class PhabricatorCursorPagedPolicyAwareQuery
       return array();
     }
 
+    $op_sub = PhutilSearchQueryCompiler::OPERATOR_SUBSTRING;
+
     $engine = $this->ferretEngine;
     $ngram_engine = new PhabricatorNgramEngine();
+    $stemmer = new PhutilSearchStemmer();
 
     $ngram_table = $engine->newNgramsObject();
     $ngram_table_name = $ngram_table->getTableName();
@@ -1422,22 +1425,49 @@ abstract class PhabricatorCursorPagedPolicyAwareQuery
 
       $length = count(phutil_utf8v($value));
 
-      if ($length >= 3) {
-        $ngrams = $ngram_engine->getNgramsFromString($value, 'query');
-        $prefix = false;
-      } else if ($length == 2) {
-        $ngrams = $ngram_engine->getNgramsFromString($value, 'prefix');
-        $prefix = false;
+      if ($raw_token->getOperator() == $op_sub) {
+        $is_substring = true;
       } else {
-        $ngrams = array(' '.$value);
-        $prefix = true;
+        $is_substring = false;
+      }
+
+      // If the user specified a substring query for a substring which is
+      // shorter than the ngram length, we can't use the ngram index, so
+      // don't do a join. We'll fall back to just doing LIKE on the full
+      // corpus.
+      if ($is_substring) {
+        if ($length < 3) {
+          continue;
+        }
+      }
+
+      if ($raw_token->isQuoted()) {
+        $is_stemmed = false;
+      } else {
+        $is_stemmed = true;
+      }
+
+      if ($is_substring) {
+        $ngrams = $ngram_engine->getNgramsFromString($value, 'query');
+      } else {
+        $ngrams = $ngram_engine->getNgramsFromString($value, 'index');
+
+        // If this is a stemmed term, only look for ngrams present in both the
+        // unstemmed and stemmed variations.
+        if ($is_stemmed) {
+          $stem_value = $stemmer->stemToken($value);
+          $stem_ngrams = $ngram_engine->getNgramsFromString(
+            $stem_value,
+            'index');
+
+          $ngrams = array_intersect($ngrams, $stem_ngrams);
+        }
       }
 
       foreach ($ngrams as $ngram) {
         $flat[] = array(
           'table' => $ngram_table_name,
           'ngram' => $ngram,
-          'prefix' => $prefix,
         );
       }
     }
@@ -1472,20 +1502,9 @@ abstract class PhabricatorCursorPagedPolicyAwareQuery
     foreach ($flat as $spec) {
       $table = $spec['table'];
       $ngram = $spec['ngram'];
-      $prefix = $spec['prefix'];
 
       $alias = 'ft'.$idx++;
 
-      if ($prefix) {
-        $joins[] = qsprintf(
-          $conn,
-          'JOIN %T %T ON %T.documentID = ftdoc.id AND %T.ngram LIKE %>',
-          $table,
-          $alias,
-          $alias,
-          $alias,
-          $ngram);
-      } else {
       $joins[] = qsprintf(
         $conn,
         'JOIN %T %T ON %T.documentID = ftdoc.id AND %T.ngram = %s',
@@ -1495,7 +1514,6 @@ abstract class PhabricatorCursorPagedPolicyAwareQuery
         $alias,
         $ngram);
     }
-    }
 
     $joins[] = qsprintf(
       $conn,
@@ -1510,16 +1528,72 @@ abstract class PhabricatorCursorPagedPolicyAwareQuery
       return array();
     }
 
+    $ngram_engine = new PhabricatorNgramEngine();
+    $stemmer = new PhutilSearchStemmer();
+    $op_sub = PhutilSearchQueryCompiler::OPERATOR_SUBSTRING;
+
     $where = array();
     foreach ($this->ferretTokens as $fulltext_token) {
       $raw_token = $fulltext_token->getToken();
       $value = $raw_token->getValue();
 
+      if ($raw_token->getOperator() == $op_sub) {
+        $is_substring = true;
+      } else {
+        $is_substring = false;
+      }
+
+      // If we're doing substring search, we just match against the raw corpus
+      // and we're done.
+      if ($is_substring) {
         $where[] = qsprintf(
           $conn,
-        '(ftfield.rawCorpus LIKE %~ OR ftfield.normalCorpus LIKE %~)',
-        $value,
+          '(ftfield.rawCorpus LIKE %~)',
           $value);
+        continue;
+      }
+
+      // Otherwise, we need to match against the term corpus and the normal
+      // corpus, so that searching for "raw" does not find "strawberry".
+      if ($raw_token->isQuoted()) {
+        $is_quoted = true;
+        $is_stemmed = false;
+      } else {
+        $is_quoted = false;
+        $is_stemmed = true;
+      }
+
+      $term_constraints = array();
+
+      $term_value = ' '.$ngram_engine->newTermsCorpus($value).' ';
+      $term_constraints[] = qsprintf(
+        $conn,
+        '(ftfield.termCorpus LIKE %~)',
+        $term_value);
+
+      if ($is_stemmed) {
+        $stem_value = $stemmer->stemToken($value);
+        $stem_value = $ngram_engine->newTermsCorpus($stem_value);
+        $stem_value = ' '.$stem_value.' ';
+
+        $term_constraints[] = qsprintf(
+          $conn,
+          '(ftfield.normalCorpus LIKE %~)',
+          $stem_value);
+      }
+
+      if ($is_quoted) {
+        $where[] = qsprintf(
+          $conn,
+          '(ftfield.rawCorpus LIKE %~ AND (%Q))',
+          $value,
+          implode(' OR ', $term_constraints));
+      } else {
+        $where[] = qsprintf(
+          $conn,
+          '(%Q)',
+          implode(' OR ', $term_constraints));
+      }
     }
 
     return $where;