Fix #8007 (Search performance on multiple words)

src/calibre/library/caches.py

@@ -181,7 +181,7 @@ class ResultCache(SearchQueryParser): # {{{
         self.search_restriction = ''
         self.field_metadata = field_metadata
         self.all_search_locations = field_metadata.get_search_terms()
-        SearchQueryParser.__init__(self, self.all_search_locations)
+        SearchQueryParser.__init__(self, self.all_search_locations, optimize=True)
         self.build_date_relop_dict()
         self.build_numeric_relop_dict()
 
@@ -264,7 +264,7 @@ class ResultCache(SearchQueryParser): # {{{
                             '<=':[2, relop_le]
                         }
 
-    def get_dates_matches(self, location, query):
+    def get_dates_matches(self, location, query, candidates):
         matches = set([])
         if len(query) < 2:
             return matches
@@ -274,13 +274,15 @@ class ResultCache(SearchQueryParser): # {{{
         loc = self.field_metadata[location]['rec_index']
 
         if query == 'false':
-            for item in self._data:
+            for id_ in candidates:
+                item = self._data[id_]
                 if item is None: continue
                 if item[loc] is None or item[loc] <= UNDEFINED_DATE:
                     matches.add(item[0])
             return matches
         if query == 'true':
-            for item in self._data:
+            for id_ in candidates:
+                item = self._data[id_]
                 if item is None: continue
                 if item[loc] is not None and item[loc] > UNDEFINED_DATE:
                     matches.add(item[0])
@@ -319,7 +321,8 @@ class ResultCache(SearchQueryParser): # {{{
                 field_count = query.count('-') + 1
             else:
                 field_count = query.count('/') + 1
-        for item in self._data:
+        for id_ in candidates:
+            item = self._data[id_]
             if item is None or item[loc] is None: continue
             if relop(item[loc], qd, field_count):
                 matches.add(item[0])
@@ -335,7 +338,7 @@ class ResultCache(SearchQueryParser): # {{{
                         '<=':[2, lambda r, q: r <= q]
                     }
 
-    def get_numeric_matches(self, location, query, val_func = None):
+    def get_numeric_matches(self, location, query, candidates, val_func = None):
         matches = set([])
         if len(query) == 0:
             return matches
@@ -381,7 +384,8 @@ class ResultCache(SearchQueryParser): # {{{
         except:
             return matches
 
-        for item in self._data:
+        for id_ in candidates:
+            item = self._data[id_]
             if item is None:
                 continue
             v = val_func(item)
@@ -393,8 +397,13 @@ class ResultCache(SearchQueryParser): # {{{
                 matches.add(item[0])
         return matches
 
-    def get_matches(self, location, query, allow_recursion=True):
+    def get_matches(self, location, query, allow_recursion=True, candidates=None):
         matches = set([])
+        if candidates is None:
+            candidates = self.universal_set()
+        if len(candidates) == 0:
+            return matches
+
         if query and query.strip():
             # get metadata key associated with the search term. Eliminates
             # dealing with plurals and other aliases
@@ -476,7 +485,8 @@ class ResultCache(SearchQueryParser): # {{{
                 else:
                     q = query
 
-                for item in self._data:
+                for id_ in candidates:
+                    item = self._data[id]
                     if item is None: continue
 
                     if col_datatype[loc] == 'bool': # complexity caused by the two-/three-value tweak

src/calibre/utils/search_query_parser.py

@@ -118,8 +118,9 @@ class SearchQueryParser(object):
                 failed.append(test[0])
         return failed
 
-    def __init__(self, locations, test=False):
+    def __init__(self, locations, test=False, optimize=False):
         self._tests_failed = False
+        self.optimize = optimize
         # Define a token
         standard_locations = map(lambda x : CaselessLiteral(x)+Suppress(':'),
                 locations)
@@ -182,38 +183,52 @@ class SearchQueryParser(object):
         # empty the list of searches used for recursion testing
         self.recurse_level = 0
         self.searches_seen = set([])
-        return self._parse(query)
+        candidates = self.universal_set()
+        return self._parse(query, candidates)
 
     # this parse is used internally because it doesn't clear the
     # recursive search test list. However, we permit seeing the
     # same search a few times because the search might appear within
     # another search.
-    def _parse(self, query):
+    def _parse(self, query, candidates=None):
         self.recurse_level += 1
         res = self._parser.parseString(query)[0]
-        t = self.evaluate(res)
+        if candidates is None:
+            candidates = self.universal_set()
+        t = self.evaluate(res, candidates)
         self.recurse_level -= 1
         return t
 
     def method(self, group_name):
         return getattr(self, 'evaluate_'+group_name)
 
-    def evaluate(self, parse_result):
+    def evaluate(self, parse_result, candidates):
-        return self.method(parse_result.getName())(parse_result)
+        return self.method(parse_result.getName())(parse_result, candidates)
 
-    def evaluate_and(self, argument):
+    def evaluate_and(self, argument, candidates):
-        return self.evaluate(argument[0]).intersection(self.evaluate(argument[1]))
+        # RHS checks only those items matched by LHS
+        # returns result of RHS check: RHmatches(LHmatches(c))
+        #  return self.evaluate(argument[0]).intersection(self.evaluate(argument[1]))
+        l = self.evaluate(argument[0], candidates)
+        return l.intersection(self.evaluate(argument[1], l))
 
-    def evaluate_or(self, argument):
+    def evaluate_or(self, argument, candidates):
-        return self.evaluate(argument[0]).union(self.evaluate(argument[1]))
+        # RHS checks only those elements not matched by LHS
+        # returns LHS union RHS: LHmatches(c) + RHmatches(c-LHmatches(c))
+        #  return self.evaluate(argument[0]).union(self.evaluate(argument[1]))
+        l = self.evaluate(argument[0], candidates)
+        return l.union(self.evaluate(argument[1], candidates.difference(l)))
 
-    def evaluate_not(self, argument):
+    def evaluate_not(self, argument, candidates):
-        return self.universal_set().difference(self.evaluate(argument[0]))
+        # unary op checks only candidates. Result: list of items matching
+        # returns: c - matches(c)
+        #  return self.universal_set().difference(self.evaluate(argument[0]))
+        return candidates.difference(self.evaluate(argument[0], candidates))
 
-    def evaluate_parenthesis(self, argument):
+    def evaluate_parenthesis(self, argument, candidates):
-        return self.evaluate(argument[0])
+        return self.evaluate(argument[0], candidates)
 
-    def evaluate_token(self, argument):
+    def evaluate_token(self, argument, candidates):
         location = argument[0]
         query = argument[1]
         if location.lower() == 'search':
@@ -224,17 +239,27 @@ class SearchQueryParser(object):
                     raise ParseException(query, len(query), 'undefined saved search', self)
                 if self.recurse_level > 5:
                     self.searches_seen.add(query)
-                return self._parse(saved_searches().lookup(query))
+                return self._parse(saved_searches().lookup(query), candidates)
             except: # convert all exceptions (e.g., missing key) to a parse error
                 raise ParseException(query, len(query), 'undefined saved search', self)
-        return self.get_matches(location, query)
+        return self._get_matches(location, query, candidates)
 
-    def get_matches(self, location, query):
+    def _get_matches(self, location, query, candidates):
+        if self.optimize:
+            return self.get_matches(location, query, candidates=candidates)
+        else:
+            return self.get_matches(location, query)
+
+    def get_matches(self, location, query, candidates=None):
         '''
         Should return the set of matches for :param:'location` and :param:`query`.
 
+        The search must be performed over all entries is :param:`candidates` is
+        None otherwise only over the items in candidates.
+
         :param:`location` is one of the items in :member:`SearchQueryParser.DEFAULT_LOCATIONS`.
         :param:`query` is a string literal.
+        :param: None or a subset of the set returned by :meth:`universal_set`.
         '''
         return set([])
 
@@ -561,7 +586,7 @@ class Tester(SearchQueryParser):
     def universal_set(self):
         return self._universal_set
 
-    def get_matches(self, location, query):
+    def get_matches(self, location, query, candidates=None):
         location = location.lower()
         if location in self.fields.keys():
             getter = operator.itemgetter(self.fields[location])
@@ -573,8 +598,13 @@ class Tester(SearchQueryParser):
         if not query:
             return set([])
         query = query.lower()
-        return set(key for key, val in self.texts.items() \
+        if candidates:
-            if query and query in getattr(getter(val), 'lower', lambda : '')())
+            return set(key for key, val in self.texts.items() \
+                if key in candidates and query and query
+                        in getattr(getter(val), 'lower', lambda : '')())
+        else:
+            return set(key for key, val in self.texts.items() \
+                if query and query in getattr(getter(val), 'lower', lambda : '')())
 
 
 
@@ -592,6 +622,7 @@ class Tester(SearchQueryParser):
 
 
 def main(args=sys.argv):
+    print 'testing unoptimized'
     tester = Tester(['authors', 'author', 'series', 'formats', 'format',
         'publisher', 'rating', 'tags', 'tag', 'comments', 'comment', 'cover',
         'isbn', 'ondevice', 'pubdate', 'size', 'date', 'title', u'#read',
@@ -601,6 +632,16 @@ def main(args=sys.argv):
         print '>>>>>>>>>>>>>> Tests Failed <<<<<<<<<<<<<<<'
         return 1
 
+    print '\n\ntesting optimized'
+    tester = Tester(['authors', 'author', 'series', 'formats', 'format',
+        'publisher', 'rating', 'tags', 'tag', 'comments', 'comment', 'cover',
+        'isbn', 'ondevice', 'pubdate', 'size', 'date', 'title', u'#read',
+        'all', 'search'], test=True, optimize=True)
+    failed = tester.run_tests()
+    if tester._tests_failed or failed:
+        print '>>>>>>>>>>>>>> Tests Failed <<<<<<<<<<<<<<<'
+        return 1
+
     return 0
 
 if __name__ == '__main__':