[pick](Variant) variant fallthrough with inverted index (#40070)

backport #40069

be/src/olap/rowset/segment_v2/segment_iterator.cpp

@@ -1412,10 +1412,15 @@ Status SegmentIterator::_init_inverted_index_iterators() {
     }
     for (auto cid : _schema->column_ids()) {
         if (_inverted_index_iterators[cid] == nullptr) {
+            // Not check type valid, since we need to get inverted index for related variant type when reading the segment.
+            // If check type valid, we can not get inverted index for variant type, and result nullptr.The result for calling
+            // get_inverted_index with variant suffix should return corresponding inverted index meta.
+            bool check_inverted_index_by_type = false;
             // Use segment’s own index_meta, for compatibility with future indexing needs to default to lowercase.
             RETURN_IF_ERROR(_segment->new_inverted_index_iterator(
                     _opts.tablet_schema->column(cid),
-                    _segment->_tablet_schema->get_inverted_index(_opts.tablet_schema->column(cid)),
+                    _segment->_tablet_schema->get_inverted_index(_opts.tablet_schema->column(cid),
+                                                                 check_inverted_index_by_type),
                     _opts, &_inverted_index_iterators[cid]));
         }
     }

regression-test/suites/variant_github_events_p2/load.groovy

@@ -169,7 +169,7 @@ suite("regression_test_variant_github_events_p2", "nonConcurrent,p2"){
 
     // build inverted index at middle of loading the data
     // ADD INDEX
-    sql """ ALTER TABLE github_events ADD INDEX idx_var (`v`) USING INVERTED PROPERTIES("parser" = "chinese", "parser_mode" = "fine_grained", "support_phrase" = "true") """
+    sql """ ALTER TABLE github_events ADD INDEX idx_var (`v`) USING INVERTED PROPERTIES("parser" = "english", "support_phrase" = "true") """
     wait_for_latest_op_on_table_finish("github_events", timeout)
 
     // 2022
@@ -212,9 +212,11 @@ suite("regression_test_variant_github_events_p2", "nonConcurrent,p2"){
         } while (running)
     }
 
-    
-    // TODO fix compaction issue, this case could be stable
+    sql """set enable_match_without_inverted_index = false""" 
+    // filter by bloom filter
     qt_sql """select cast(v["payload"]["pull_request"]["additions"] as int)  from github_events where cast(v["repo"]["name"] as string) = 'xpressengine/xe-core' order by 1;"""
     qt_sql """select * from github_events where  cast(v["repo"]["name"] as string) = 'xpressengine/xe-core' order by 1 limit 10"""
-    // TODO add test case that some certain columns are materialized in some file while others are not materilized(sparse)
+    // query with inverted index
+    qt_sql """select cast(v["payload"]["pull_request"]["additions"] as int)  from github_events where v["repo"]["name"] match 'xpressengine' order by 1;"""
+    qt_sql """select count()  from github_events where v["repo"]["name"] match 'apache' order by 1;"""
 }

regression-test/suites/variant_p0/with_index/load.groovy

@@ -61,6 +61,7 @@ suite("regression_test_variant_with_index", "nonConcurrent"){
         properties("replication_num" = "1", "disable_auto_compaction" = "true");
     """
     sql """insert into var_with_index values(1, '{"a" : 0, "b": 3}', 'hello world'), (2, '{"a" : 123}', 'world'),(3, '{"a" : 123}', 'hello world')"""
+    sql """set enable_match_without_inverted_index = false""" 
     qt_sql_inv_1 """select v["a"] from var_with_index where inv match 'hello' order by k"""
     qt_sql_inv_2 """select v["a"] from var_with_index where inv match 'hello' and cast(v['a'] as int) > 0 order by k"""
     qt_sql_inv_3 """select * from var_with_index where inv match 'hello' and cast(v["a"] as int) > 0 order by k"""

regression-test/suites/variant_p0/with_index/var_index.groovy

@@ -33,7 +33,9 @@ suite("regression_test_variant_var_index", "p0"){
     sql """insert into var_index values(2, '{"a" : 18811, "b" : "hello world", "c" : 1181111}')"""
     sql """insert into var_index values(3, '{"a" : 18811, "b" : "hello wworld", "c" : 11111}')"""
     sql """insert into var_index values(4, '{"a" : 1234, "b" : "hello xxx world", "c" : 8181111}')"""
+    sql """set enable_match_without_inverted_index = false""" 
     qt_sql """select * from var_index where cast(v["a"] as smallint) > 123 and cast(v["b"] as string) match 'hello' and cast(v["c"] as int) > 1024 order by k"""
+    sql """set enable_match_without_inverted_index = true""" 
     sql """insert into var_index values(5, '{"a" : 123456789, "b" : 123456, "c" : 8181111}')"""
     qt_sql """select * from var_index where cast(v["a"] as int) > 123 and cast(v["b"] as string) match 'hello' and cast(v["c"] as int) > 11111 order by k"""
     // insert double/float/array/json