[Fix](Variant) check storage type before apply get_row_ranges_by_dict

    1. storage type maybe missmatch with predicate type eg. cast(v['a'] as date) = '2020-01-01'. v['a'] in storage is string type, but predicate type is date type

    2. serialization object with empty subcolumns may lost num_rows, so need to record num_rows and set back num_rows in serdes

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

@@ -699,6 +699,11 @@ Status SegmentIterator::_get_row_ranges_from_conditions(RowRanges* condition_row
         if (_opts.io_ctx.reader_type == ReaderType::READER_QUERY) {
             RowRanges dict_row_ranges = RowRanges::create_single(num_rows());
             for (auto cid : cids) {
+                if (!_segment->can_apply_predicate_safely(cid,
+                                                          _opts.col_id_to_predicates.at(cid).get(),
+                                                          *_schema, _opts.io_ctx.reader_type)) {
+                    continue;
+                }
                 RowRanges tmp_row_ranges = RowRanges::create_single(num_rows());
                 DCHECK(_opts.col_id_to_predicates.count(cid) > 0);
                 RETURN_IF_ERROR(_column_iterators[cid]->get_row_ranges_by_dict(

be/src/vec/data_types/data_type_object.cpp

@@ -61,12 +61,13 @@ int64_t DataTypeObject::get_uncompressed_serialized_bytes(const IColumn& column,
     size_t size = 0;
 
     size += sizeof(uint32_t);
+    size_t num_of_columns = 0;
     for (const auto& entry : subcolumns) {
         auto type = entry->data.get_least_common_type();
         if (is_nothing(type)) {
             continue;
         }
-
+        ++num_of_columns;
         PColumnMeta column_meta_pb;
         column_meta_pb.set_name(entry->path.get_path());
         type->to_pb_column_meta(&column_meta_pb);
@@ -78,6 +79,10 @@ int64_t DataTypeObject::get_uncompressed_serialized_bytes(const IColumn& column,
         size += type->get_uncompressed_serialized_bytes(entry->data.get_finalized_column(),
                                                         be_exec_version);
     }
+    if (num_of_columns == 0) {
+        // Make sure the num_of_rows in ColumnObject not lost
+        size += sizeof(uint32_t);
+    }
 
     return size;
 }
@@ -121,6 +126,11 @@ char* DataTypeObject::serialize(const IColumn& column, char* buf, int be_exec_ve
     }
     // serialize num of subcolumns
     *reinterpret_cast<uint32_t*>(size_pos) = num_of_columns;
+    if (num_of_columns == 0) {
+        // Make sure the num_of_rows in ColumnObject not lost
+        *reinterpret_cast<uint32_t*>(buf) = column_object.rows();
+        buf += sizeof(uint32_t);
+    }
 
     return buf;
 }
@@ -155,6 +165,13 @@ const char* DataTypeObject::deserialize(const char* buf, IColumn* column,
         }
         column_object->add_sub_column(key, std::move(sub_column), type);
     }
+    size_t num_rows = 0;
+    if (num_subcolumns == 0) {
+        // Make sure the num_of_rows in ColumnObject not lost
+        num_rows = *reinterpret_cast<const uint32_t*>(buf);
+        column_object->set_num_rows(num_rows);
+        buf += sizeof(uint32_t);
+    }
 
     column_object->finalize();
 #ifndef NDEBUG

regression-test/data/variant_p0/rqg/rqg4.out

@@ -0,0 +1,16 @@
+-- This file is automatically generated. You should know what you did if you want to edit this
+-- !rqg4 --
+0
+
+-- !rqg4_2 --
+500
+
+-- !rqg4_3 --
+500
+
+-- !rqg4_4 --
+70
+
+-- !rqg4_5 --
+70
+

regression-test/data/variant_p0/rqg/rqg5.out

@@ -0,0 +1,67 @@
+-- This file is automatically generated. You should know what you did if you want to edit this
+-- !rqg5 --
+0
+
+-- !rqg5_2 --
+50
+
+-- !rqg5_3 --
+50
+
+-- !rqg5_4 --
+50
+
+-- !rqg5_5 --
+50
+
+-- !rqg5_6 --
+50
+
+-- !rqg5_7 --
+0
+
+-- !rqg5_8 --
+0
+
+-- !rqg5_9 --
+50
+
+-- !rqg5_10 --
+50
+
+-- !rqg5_11 --
+50
+
+-- !rqg5_12 --
+50
+
+-- !rqg5_13 --
+50
+
+-- !rqg5_14 --
+50
+
+-- !rqg5_15 --
+0
+
+-- !rqg5_16 --
+100
+
+-- !rqg5_17 --
+100
+
+-- !rqg5_18 --
+100
+
+-- !rqg5_19 --
+100
+
+-- !rqg5_20 --
+100
+
+-- !rqg5_21 --
+100
+
+-- !rqg5_22 --
+0
+