This is an automated email from the ASF dual-hosted git repository.

eldenmoon pushed a commit to branch branch-2.0-var
in repository https://gitbox.apache.org/repos/asf/doris.git


The following commit(s) were added to refs/heads/branch-2.0-var by this push:
     new 361ae5466ed [performance](variant) support topn 2phase read for 
variant column (#28381)
361ae5466ed is described below

commit 361ae5466ed4b5fefa497a6d75e865655318814e
Author: lihangyu <[email protected]>
AuthorDate: Fri Dec 15 19:56:47 2023 +0800

    [performance](variant) support topn 2phase read for variant column (#28381)
    
    * [performance](variant) support topn 2phase read for variant column
    
    * fix segment use after free
    
    * fix use after free
    
    * add case
    
    * fix case
    
    * fix
---
 .../rowset/segment_v2/hierarchical_data_reader.h   |  4 +-
 be/src/olap/rowset/segment_v2/segment.cpp          | 75 ++++++++++++++++++
 be/src/olap/rowset/segment_v2/segment.h            | 11 +++
 be/src/olap/tablet_schema.cpp                      | 14 ++++
 be/src/olap/tablet_schema.h                        |  5 ++
 be/src/runtime/descriptors.cpp                     |  6 +-
 be/src/service/internal_service.cpp                | 90 +++++++++++++++-------
 be/src/vec/columns/column_object.cpp               |  9 ++-
 be/src/vec/exec/scan/new_olap_scanner.cpp          | 40 ++++------
 be/src/vec/json/path_in_data.cpp                   | 11 +++
 be/src/vec/json/path_in_data.h                     |  1 +
 .../java/org/apache/doris/analysis/SelectStmt.java |  3 +-
 gensrc/proto/descriptors.proto                     |  1 +
 regression-test/data/variant_p0/sql/gh_data.out    | 86 +++++++++++----------
 .../suites/variant_p0/column_size_limit.groovy     | 18 ++++-
 regression-test/suites/variant_p0/load.groovy      |  6 +-
 regression-test/suites/variant_p0/sql/gh_data.sql  | 11 ++-
 17 files changed, 276 insertions(+), 115 deletions(-)

diff --git a/be/src/olap/rowset/segment_v2/hierarchical_data_reader.h 
b/be/src/olap/rowset/segment_v2/hierarchical_data_reader.h
index a56e80aa464..99bbb36ab87 100644
--- a/be/src/olap/rowset/segment_v2/hierarchical_data_reader.h
+++ b/be/src/olap/rowset/segment_v2/hierarchical_data_reader.h
@@ -140,7 +140,7 @@ private:
 
         // copy container variant to dst variant, todo avoid copy
         variant.insert_range_from(container_variant, 0, nrows);
-        variant.set_num_rows(nrows);
+        // variant.set_num_rows(nrows);
         _rows_read += nrows;
         variant.finalize();
 #ifndef NDEBUG
@@ -180,7 +180,7 @@ public:
 private:
     Status extract_to(vectorized::MutableColumnPtr& dst, size_t nrows);
 
-    const TabletColumn& _col;
+    TabletColumn _col;
     // may shared among different column iterators
     std::unique_ptr<StreamReader> _root_reader;
 };
diff --git a/be/src/olap/rowset/segment_v2/segment.cpp 
b/be/src/olap/rowset/segment_v2/segment.cpp
index f1285977047..9c12fc0e074 100644
--- a/be/src/olap/rowset/segment_v2/segment.cpp
+++ b/be/src/olap/rowset/segment_v2/segment.cpp
@@ -60,10 +60,12 @@
 #include "util/slice.h" // Slice
 #include "vec/columns/column.h"
 #include "vec/common/string_ref.h"
+#include "vec/core/field.h"
 #include "vec/data_types/data_type.h"
 #include "vec/data_types/data_type_factory.hpp"
 #include "vec/data_types/data_type_nullable.h"
 #include "vec/data_types/data_type_object.h"
+#include "vec/json/path_in_data.h"
 #include "vec/olap/vgeneric_iterators.h"
 
 namespace doris {
@@ -672,5 +674,78 @@ bool Segment::is_same_file_col_type_with_expected(int32_t 
cid, const Schema& sch
     return (!file_column_type) || (file_column_type && 
file_column_type->equals(*expected_type));
 }
 
+ std::shared_ptr<const vectorized::IDataType>
+ Segment::get_data_type_of(vectorized::PathInData path,
+                            bool ignore_children) const {
+    auto node = _sub_column_tree.find_leaf(path);
+    if (node) {
+        if (ignore_children || node->children.empty()) {
+            return node->data.file_column_type;
+        }
+    }
+    // it contains children or column missing in storage, so treat it as 
variant
+    return std::make_shared<vectorized::DataTypeObject>();
+}
+
+Status Segment::seek_and_read_by_rowid(const TabletSchema& schema, 
SlotDescriptor* slot,
+                                       uint32_t row_id, 
vectorized::MutableColumnPtr& result,
+                                       OlapReaderStatistics& stats,
+                                       std::unique_ptr<ColumnIterator>& 
iterator_hint) {
+    StorageReadOptions storage_read_opt;
+    storage_read_opt.io_ctx.reader_type = ReaderType::READER_QUERY;
+    io::IOContext io_ctx;
+    io_ctx.reader_type = ReaderType::READER_QUERY;
+    segment_v2::ColumnIteratorOptions opt {
+            .file_reader = file_reader().get(),
+            .stats = &stats,
+            .use_page_cache = !config::disable_storage_page_cache,
+            .io_ctx = io_ctx,
+    };
+    std::vector<segment_v2::rowid_t> single_row_loc {row_id};
+    if (!slot->column_paths().empty()) {
+        vectorized::PathInData 
path(schema.column_by_uid(slot->col_unique_id()).name_lower_case(),
+                                    slot->column_paths());
+        auto storage_type = get_data_type_of(path, false);
+        vectorized::MutableColumnPtr file_storage_column = 
storage_type->create_column();
+        DCHECK(storage_type != nullptr);
+        TabletColumn column = TabletColumn::create_materialized_variant_column(
+                schema.column_by_uid(slot->col_unique_id()).name_lower_case(), 
slot->column_paths(),
+                slot->col_unique_id());
+        if (iterator_hint == nullptr) {
+            RETURN_IF_ERROR(
+                    new_column_iterator(column, &iterator_hint, 
&storage_read_opt));
+            RETURN_IF_ERROR(iterator_hint->init(opt));
+        }
+        RETURN_IF_ERROR(
+                iterator_hint->read_by_rowids(single_row_loc.data(), 1, 
file_storage_column));
+        // iterator_hint.reset(nullptr);
+        // Get it's inner field, for JSONB case
+        vectorized::Field field = remove_nullable(storage_type)->get_default();
+        file_storage_column->get(0, field);
+        result->insert(field);
+    } else {
+        int index = -1;
+        if (slot->col_unique_id() >= 0) {
+            index = schema.field_index(slot->col_name());
+        } else {
+            index = schema.field_index(slot->col_name());
+        }
+        if (index < 0) {
+            std::stringstream ss;
+            ss << "field name is invalid. field=" << slot->col_name()
+               << ", field_name_to_index=" << schema.get_all_field_names();
+            return Status::InternalError(ss.str());
+        }
+        storage_read_opt.io_ctx.reader_type = ReaderType::READER_QUERY;
+        if (iterator_hint == nullptr) {
+            RETURN_IF_ERROR(
+                    new_column_iterator(schema.column(index), &iterator_hint, 
&storage_read_opt));
+            RETURN_IF_ERROR(iterator_hint->init(opt));
+        }
+        RETURN_IF_ERROR(iterator_hint->read_by_rowids(single_row_loc.data(), 
1, result));
+    }
+    return Status::OK();
+}
+
 } // namespace segment_v2
 } // namespace doris
diff --git a/be/src/olap/rowset/segment_v2/segment.h 
b/be/src/olap/rowset/segment_v2/segment.h
index 7983f86b3a8..f504d624a96 100644
--- a/be/src/olap/rowset/segment_v2/segment.h
+++ b/be/src/olap/rowset/segment_v2/segment.h
@@ -39,9 +39,14 @@
 #include "olap/rowset/segment_v2/page_handle.h"
 #include "olap/schema.h"
 #include "olap/tablet_schema.h"
+#include "runtime/descriptors.h"
 #include "util/once.h"
 #include "util/slice.h"
+#include "vec/columns/column.h"
 #include "vec/columns/subcolumn_tree.h"
+#include "vec/data_types/data_type.h"
+#include "vec/data_types/data_type_nullable.h"
+#include "vec/json/path_in_data.h"
 
 namespace doris {
 namespace vectorized {
@@ -124,6 +129,10 @@ public:
 
     Status read_key_by_rowid(uint32_t row_id, std::string* key);
 
+    Status seek_and_read_by_rowid(const TabletSchema& schema, SlotDescriptor* 
slot, uint32_t row_id,
+                                  vectorized::MutableColumnPtr& result, 
OlapReaderStatistics& stats,
+                                  std::unique_ptr<ColumnIterator>& 
iterator_hint);
+
     Status load_index();
 
     Status load_pk_index_and_bf();
@@ -147,6 +156,8 @@ public:
     std::shared_ptr<const vectorized::IDataType> get_data_type_of(const Field& 
filed,
                                                                   bool 
ignore_children) const;
 
+    std::shared_ptr<const vectorized::IDataType> 
get_data_type_of(vectorized::PathInData path,
+                                                                  bool 
ignore_children) const;
     // If column in segment is the same type in schema
     bool is_same_file_col_type_with_expected(int32_t cid, const Schema& schema,
                                              bool ignore_children) const;
diff --git a/be/src/olap/tablet_schema.cpp b/be/src/olap/tablet_schema.cpp
index 0191241e5d3..c227b9cc06a 100644
--- a/be/src/olap/tablet_schema.cpp
+++ b/be/src/olap/tablet_schema.cpp
@@ -457,6 +457,20 @@ void TabletColumn::init_from_pb(const ColumnPB& column) {
     }
 }
 
+TabletColumn TabletColumn::create_materialized_variant_column(const 
std::string& root,
+                                                              const 
std::vector<std::string>& paths,
+                                                              int32_t 
parent_unique_id) {
+    TabletColumn subcol;
+    subcol.set_type(FieldType::OLAP_FIELD_TYPE_VARIANT);
+    subcol.set_is_nullable(true);
+    subcol.set_unique_id(-1);
+    subcol.set_parent_unique_id(parent_unique_id);
+    vectorized::PathInData path(root, paths);
+    subcol.set_path_info(path);
+    subcol.set_name(path.get_path());
+    return subcol;
+}
+
 void TabletColumn::to_schema_pb(ColumnPB* column) const {
     column->set_unique_id(_unique_id);
     column->set_name(_col_name);
diff --git a/be/src/olap/tablet_schema.h b/be/src/olap/tablet_schema.h
index 087fb437eca..79401d0d6b8 100644
--- a/be/src/olap/tablet_schema.h
+++ b/be/src/olap/tablet_schema.h
@@ -88,6 +88,11 @@ public:
                _type == FieldType::OLAP_FIELD_TYPE_QUANTILE_STATE ||
                _type == FieldType::OLAP_FIELD_TYPE_AGG_STATE;
     }
+    // Such columns are not exist in frontend schema info, so we need to
+    // add them into tablet_schema for later column indexing.
+    static TabletColumn create_materialized_variant_column(const std::string& 
root,
+                                                           const 
std::vector<std::string>& paths,
+                                                           int32_t 
parent_unique_id);
     bool has_default_value() const { return _has_default_value; }
     std::string default_value() const { return _default_value; }
     size_t length() const { return _length; }
diff --git a/be/src/runtime/descriptors.cpp b/be/src/runtime/descriptors.cpp
index f8125588ae5..435df6648c6 100644
--- a/be/src/runtime/descriptors.cpp
+++ b/be/src/runtime/descriptors.cpp
@@ -83,7 +83,8 @@ SlotDescriptor::SlotDescriptor(const PSlotDescriptor& pdesc)
           _field_idx(-1),
           _is_materialized(pdesc.is_materialized()),
           _is_key(pdesc.is_key()),
-          _need_materialize(true) {}
+          _need_materialize(true),
+          _column_paths(pdesc.column_paths().begin(), 
pdesc.column_paths().end()) {}
 
 void SlotDescriptor::to_protobuf(PSlotDescriptor* pslot) const {
     pslot->set_id(_id);
@@ -100,6 +101,9 @@ void SlotDescriptor::to_protobuf(PSlotDescriptor* pslot) 
const {
     pslot->set_col_unique_id(_col_unique_id);
     pslot->set_is_key(_is_key);
     pslot->set_col_type(_col_type);
+    for (const std::string& path : _column_paths) {
+        pslot->add_column_paths(path);
+    }
 }
 
 vectorized::MutableColumnPtr SlotDescriptor::get_empty_mutable_column() const {
diff --git a/be/src/service/internal_service.cpp 
b/be/src/service/internal_service.cpp
index 597c8fdfef4..2a69400fe3f 100644
--- a/be/src/service/internal_service.cpp
+++ b/be/src/service/internal_service.cpp
@@ -25,6 +25,7 @@
 #include <butil/errno.h>
 #include <butil/iobuf.h>
 #include <fcntl.h>
+#include <fmt/core.h>
 #include <gen_cpp/MasterService_types.h>
 #include <gen_cpp/PaloInternalService_types.h>
 #include <gen_cpp/PlanNodes_types.h>
@@ -46,6 +47,7 @@
 #include <set>
 #include <sstream>
 #include <string>
+#include <unordered_map>
 #include <utility>
 #include <vector>
 
@@ -69,6 +71,7 @@
 #include "olap/rowset/segment_v2/common.h"
 #include "olap/rowset/segment_v2/inverted_index_desc.h"
 #include "olap/rowset/segment_v2/segment.h"
+#include "olap/rowset/segment_v2/segment_iterator.h"
 #include "olap/segment_loader.h"
 #include "olap/storage_engine.h"
 #include "olap/tablet.h"
@@ -1596,6 +1599,38 @@ auto scope_timer_run(Func fn, int64_t* cost) -> 
decltype(fn()) {
     return res;
 }
 
+struct IteratorKey {
+    int64_t tablet_id;
+    RowsetId rowset_id;
+    uint64_t segment_id;
+    int slot_id;
+
+    // unordered map std::equal_to
+    bool operator==(const IteratorKey& rhs) const {
+        return tablet_id == rhs.tablet_id && rowset_id == rhs.rowset_id &&
+               segment_id == rhs.segment_id && slot_id == rhs.slot_id;
+    }
+};
+
+struct HashOfIteratorKey {
+    size_t operator()(const IteratorKey& key) const {
+        size_t seed = 0;
+        seed = HashUtil::hash64(&key.tablet_id, sizeof(key.tablet_id), seed);
+        seed = HashUtil::hash64(&key.rowset_id.hi, sizeof(key.rowset_id.hi), 
seed);
+        seed = HashUtil::hash64(&key.rowset_id.mi, sizeof(key.rowset_id.mi), 
seed);
+        seed = HashUtil::hash64(&key.rowset_id.lo, sizeof(key.rowset_id.lo), 
seed);
+        seed = HashUtil::hash64(&key.segment_id, sizeof(key.segment_id), seed);
+        seed = HashUtil::hash64(&key.slot_id, sizeof(key.slot_id), seed);
+        return seed;
+    }
+};
+
+struct IteratorItem {
+    std::unique_ptr<ColumnIterator> iterator;
+    // for holding the reference of segment to avoid use after release
+    SegmentSharedPtr segment;
+};
+
 Status PInternalServiceImpl::_multi_get(const PMultiGetRequest& request,
                                         PMultiGetResponse* response) {
     OlapReaderStatistics stats;
@@ -1620,6 +1655,7 @@ Status PInternalServiceImpl::_multi_get(const 
PMultiGetRequest& request,
         full_read_schema.append_column(TabletColumn(column_pb));
     }
 
+    std::unordered_map<IteratorKey, IteratorItem, HashOfIteratorKey> 
iterator_map;
     // read row by row
     for (size_t i = 0; i < request.row_locs_size(); ++i) {
         const auto& row_loc = request.row_locs(i);
@@ -1658,8 +1694,8 @@ Status PInternalServiceImpl::_multi_get(const 
PMultiGetRequest& request,
                 },
                 &acquire_segments_ms));
         // find segment
-        auto it = std::find_if(segment_cache.get_segments().begin(),
-                               segment_cache.get_segments().end(),
+        auto it = std::find_if(segment_cache.get_segments().cbegin(),
+                               segment_cache.get_segments().cend(),
                                [&row_loc](const segment_v2::SegmentSharedPtr& 
seg) {
                                    return seg->id() == row_loc.segment_id();
                                });
@@ -1687,33 +1723,28 @@ Status PInternalServiceImpl::_multi_get(const 
PMultiGetRequest& request,
         if (result_block.is_empty_column()) {
             result_block = vectorized::Block(desc.slots(), 
request.row_locs().size());
         }
+        VLOG_DEBUG << "Read row location "
+                   << fmt::format("{}, {}, {}, {}", row_location.tablet_id,
+                                  
row_location.row_location.rowset_id.to_string(),
+                                  row_location.row_location.segment_id,
+                                  row_location.row_location.row_id);
         for (int x = 0; x < desc.slots().size(); ++x) {
-            int index = -1;
-            if (desc.slots()[x]->col_unique_id() >= 0) {
-                // light sc enabled
-                index = 
full_read_schema.field_index(desc.slots()[x]->col_unique_id());
-            } else {
-                index = 
full_read_schema.field_index(desc.slots()[x]->col_name());
-            }
-            if (index < 0) {
-                std::stringstream ss;
-                ss << "field name is invalid. field=" << 
desc.slots()[x]->col_name()
-                   << ", field_name_to_index=" << 
full_read_schema.get_all_field_names();
-                return Status::InternalError(ss.str());
-            }
-            std::unique_ptr<segment_v2::ColumnIterator> column_iterator;
+            auto row_id = 
static_cast<segment_v2::rowid_t>(row_loc.ordinal_id());
             vectorized::MutableColumnPtr column =
                     result_block.get_by_position(x).column->assume_mutable();
-            RETURN_IF_ERROR(
-                    
segment->new_column_iterator(full_read_schema.column(index), &column_iterator));
-            segment_v2::ColumnIteratorOptions opt;
-            opt.file_reader = segment->file_reader().get();
-            opt.stats = &stats;
-            opt.use_page_cache = !config::disable_storage_page_cache;
-            column_iterator->init(opt);
-            std::vector<segment_v2::rowid_t> single_row_loc {
-                    static_cast<segment_v2::rowid_t>(row_loc.ordinal_id())};
-            
RETURN_IF_ERROR(column_iterator->read_by_rowids(single_row_loc.data(), 1, 
column));
+            IteratorKey iterator_key {.tablet_id = tablet->tablet_id(),
+                                      .rowset_id = rowset_id,
+                                      .segment_id = row_loc.segment_id(),
+                                      .slot_id = desc.slots()[x]->id()};
+            IteratorItem& iterator_item = iterator_map[iterator_key];
+            if (iterator_item.segment == nullptr) {
+                // hold the reference
+                iterator_map[iterator_key].segment = segment;
+            }
+            segment = iterator_item.segment;
+            RETURN_IF_ERROR(segment->seek_and_read_by_rowid(full_read_schema, 
desc.slots()[x],
+                                                            row_id, column, 
stats,
+                                                            
iterator_item.iterator));
         }
     }
     // serialize block if not empty
@@ -1733,11 +1764,13 @@ Status PInternalServiceImpl::_multi_get(const 
PMultiGetRequest& request,
                          "hit_cached_pages:{}, total_pages_read:{}, 
compressed_bytes_read:{}, "
                          "io_latency:{}ns, "
                          "uncompressed_bytes_read:{},"
+                         "bytes_read:{},"
                          "acquire_tablet_ms:{}, acquire_rowsets_ms:{}, 
acquire_segments_ms:{}, "
                          "lookup_row_data_ms:{}",
                          stats.cached_pages_num, stats.total_pages_num, 
stats.compressed_bytes_read,
-                         stats.io_ns, stats.uncompressed_bytes_read, 
acquire_tablet_ms,
-                         acquire_rowsets_ms, acquire_segments_ms, 
lookup_row_data_ms);
+                         stats.io_ns, stats.uncompressed_bytes_read, 
stats.bytes_read,
+                         acquire_tablet_ms, acquire_rowsets_ms, 
acquire_segments_ms,
+                         lookup_row_data_ms);
     return Status::OK();
 }
 
@@ -1746,6 +1779,7 @@ void 
PInternalServiceImpl::multiget_data(google::protobuf::RpcController* contro
                                          PMultiGetResponse* response,
                                          google::protobuf::Closure* done) {
     bool ret = _light_work_pool.try_offer([request, response, done, this]() {
+        signal::set_signal_task_id(request->query_id());
         // multi get data by rowid
         MonotonicStopWatch watch;
         watch.start();
diff --git a/be/src/vec/columns/column_object.cpp 
b/be/src/vec/columns/column_object.cpp
index c8818505790..c889562d83b 100644
--- a/be/src/vec/columns/column_object.cpp
+++ b/be/src/vec/columns/column_object.cpp
@@ -642,11 +642,12 @@ void ColumnObject::for_each_subcolumn(ColumnCallback 
callback) {
 }
 
 void ColumnObject::insert_from(const IColumn& src, size_t n) {
-    const auto& src_v = assert_cast<const ColumnObject&>(src);
+    const auto* src_v = check_and_get_column<ColumnObject>(src);
     // optimize when src and this column are scalar variant, since try_insert 
is inefficiency
-    if (src_v.is_scalar_variant() && is_scalar_variant() &&
-        src_v.get_root_type()->equals(*get_root_type()) && 
src_v.is_finalized() && is_finalized()) {
-        
assert_cast<ColumnNullable&>(*get_root()).insert_from(*src_v.get_root(), n);
+    if (src_v != nullptr && src_v->is_scalar_variant() && is_scalar_variant() 
&&
+        src_v->get_root_type()->equals(*get_root_type()) && 
src_v->is_finalized() &&
+        is_finalized()) {
+        
assert_cast<ColumnNullable&>(*get_root()).insert_from(*src_v->get_root(), n);
         ++num_rows;
         return;
     }
diff --git a/be/src/vec/exec/scan/new_olap_scanner.cpp 
b/be/src/vec/exec/scan/new_olap_scanner.cpp
index d6c84b193af..8f319af81f1 100644
--- a/be/src/vec/exec/scan/new_olap_scanner.cpp
+++ b/be/src/vec/exec/scan/new_olap_scanner.cpp
@@ -56,6 +56,7 @@
 #include "vec/exec/scan/new_olap_scan_node.h"
 #include "vec/exec/scan/vscan_node.h"
 #include "vec/exprs/vexpr_context.h"
+#include "vec/json/path_in_data.h"
 #include "vec/olap/block_reader.h"
 
 namespace doris::vectorized {
@@ -404,20 +405,6 @@ Status NewOlapScanner::_init_tablet_reader_params(
     return Status::OK();
 }
 
-vectorized::PathInData NewOlapScanner::_build_path(SlotDescriptor* slot) {
-    PathInDataBuilder path_builder;
-    const std::string& col_name = slot->col_name_lower_case();
-    auto delimeter_index = col_name.find(".");
-    std::string_view root_name = delimeter_index == std::string::npos
-                                         ? col_name
-                                         : std::string_view(col_name.data(), 
delimeter_index);
-    path_builder = path_builder.append(root_name, false);
-    for (const std::string& path : slot->column_paths()) {
-        path_builder.append(path, false);
-    }
-    return path_builder.build();
-}
-
 Status NewOlapScanner::_init_variant_columns() {
     // Parent column has path info to distinction from each other
     for (auto slot : _output_tuple_desc->slots()) {
@@ -430,21 +417,18 @@ Status NewOlapScanner::_init_variant_columns() {
         if (slot->type().is_variant_type()) {
             // Such columns are not exist in frontend schema info, so we need 
to
             // add them into tablet_schema for later column indexing.
-            TabletColumn subcol;
-            subcol.set_type(FieldType::OLAP_FIELD_TYPE_VARIANT);
-            subcol.set_name(slot->col_name());
-            subcol.set_is_nullable(true);
-            subcol.set_unique_id(slot->col_unique_id());
-            PathInData path = _build_path(slot);
-            subcol.set_path_info(path);
-            if (_tablet_schema->field_index(path) < 0) {
+            TabletColumn subcol = 
TabletColumn::create_materialized_variant_column(
+                    
_tablet_schema->column_by_uid(slot->col_unique_id()).name_lower_case(),
+                    slot->column_paths(), slot->col_unique_id());
+            if (_tablet_schema->field_index(subcol.path_info()) < 0) {
                 _tablet_schema->append_column(subcol, 
TabletSchema::ColumnType::VARIANT);
             }
         } else if (!slot->column_paths().empty()) {
-            // Extracted materialized columns update it's path info
-            PathInData path = _build_path(slot);
-            int index = _tablet_schema->field_index(slot->col_unique_id());
-            _tablet_schema->mutable_columns()[index].set_path_info(path);
+            CHECK(false) << "Not implemented";
+            // // Extracted materialized columns update it's path info
+            // PathInData path = _build_path(slot);
+            // int index = _tablet_schema->field_index(slot->col_unique_id());
+            // _tablet_schema->mutable_columns()[index].set_path_info(path);
         }
     }
     return Status::OK();
@@ -462,7 +446,9 @@ Status NewOlapScanner::_init_return_columns() {
         // variant column using path to index a column
         int32_t index = 0;
         if (slot->type().is_variant_type()) {
-            index = _tablet_schema->field_index(_build_path(slot));
+            index = _tablet_schema->field_index(PathInData(
+                    
_tablet_schema->column_by_uid(slot->col_unique_id()).name_lower_case(),
+                    slot->column_paths()));
         } else {
             index = slot->col_unique_id() >= 0 ? 
_tablet_schema->field_index(slot->col_unique_id())
                                                : 
_tablet_schema->field_index(slot->col_name());
diff --git a/be/src/vec/json/path_in_data.cpp b/be/src/vec/json/path_in_data.cpp
index 0ed1975339f..c32f97d7d0a 100644
--- a/be/src/vec/json/path_in_data.cpp
+++ b/be/src/vec/json/path_in_data.cpp
@@ -46,11 +46,22 @@ PathInData::PathInData(const PathInData& other) : 
path(other.path) {
     build_parts(other.get_parts());
 }
 
+PathInData::PathInData(const std::string& root, const 
std::vector<std::string>& paths) {
+    PathInDataBuilder path_builder;
+    path_builder.append(root, false);
+    for (const std::string& path : paths) {
+        path_builder.append(path, false);
+    }
+    build_path(path_builder.get_parts());
+    build_parts(path_builder.get_parts());
+}
+
 PathInData::PathInData(const std::vector<std::string>& paths) {
     PathInDataBuilder path_builder;
     for (size_t i = 0; i < paths.size(); ++i) {
         path_builder.append(paths[i], false);
     }
+    build_path(path_builder.get_parts());
     build_parts(path_builder.get_parts());
 }
 
diff --git a/be/src/vec/json/path_in_data.h b/be/src/vec/json/path_in_data.h
index 87278c1c925..ddb27b14cf5 100644
--- a/be/src/vec/json/path_in_data.h
+++ b/be/src/vec/json/path_in_data.h
@@ -61,6 +61,7 @@ public:
     explicit PathInData(std::string_view path_);
     explicit PathInData(const Parts& parts_);
     explicit PathInData(const std::vector<std::string>& paths);
+    explicit PathInData(const std::string& root, const 
std::vector<std::string>& pathsd);
     PathInData(const PathInData& other);
     PathInData& operator=(const PathInData& other);
     static UInt128 get_parts_hash(const Parts& parts_);
diff --git a/fe/fe-core/src/main/java/org/apache/doris/analysis/SelectStmt.java 
b/fe/fe-core/src/main/java/org/apache/doris/analysis/SelectStmt.java
index 8f34046401c..fe7a804f64d 100644
--- a/fe/fe-core/src/main/java/org/apache/doris/analysis/SelectStmt.java
+++ b/fe/fe-core/src/main/java/org/apache/doris/analysis/SelectStmt.java
@@ -798,8 +798,7 @@ public class SelectStmt extends QueryStmt {
             LOG.debug("only support duplicate key or MOW model");
             return false;
         }
-        if (!olapTable.getEnableLightSchemaChange() || 
!Strings.isNullOrEmpty(olapTable.getStoragePolicy())
-                    || olapTable.hasVariantColumns()) {
+        if (!olapTable.getEnableLightSchemaChange()) {
             return false;
         }
         if (getOrderByElements() != null) {
diff --git a/gensrc/proto/descriptors.proto b/gensrc/proto/descriptors.proto
index 270199cc020..8762a78c0f4 100644
--- a/gensrc/proto/descriptors.proto
+++ b/gensrc/proto/descriptors.proto
@@ -38,6 +38,7 @@ message PSlotDescriptor {
     optional bool is_key = 12;
     optional bool is_auto_increment = 13;
     optional int32 col_type = 14 [default = 0];
+    repeated string column_paths = 15;
 };
 
 message PTupleDescriptor {
diff --git a/regression-test/data/variant_p0/sql/gh_data.out 
b/regression-test/data/variant_p0/sql/gh_data.out
index 232f4071d33..257f1fb0ce6 100644
--- a/regression-test/data/variant_p0/sql/gh_data.out
+++ b/regression-test/data/variant_p0/sql/gh_data.out
@@ -3,65 +3,47 @@
 0
 
 -- !gh_data_2 --
-5000
+0
 
 -- !gh_data_3 --
+0
+
+-- !gh_data_4 --
+5000
+
+-- !gh_data_5 --
 leonardomso/33-js-concepts     3
 ytdl-org/youtube-dl    3
 Bogdanp/neko   2
 bminossi/AllVideoPocsFromHackerOne     2
 disclose/diodata       2
 
--- !gh_data_4 --
-14690758274
-
--- !gh_data_5 --
-73453762334584
-
 -- !gh_data_6 --
-457806339
+14690758274
 
 -- !gh_data_7 --
-0
+73453762334584
 
 -- !gh_data_8 --
-19829
+457806339
 
 -- !gh_data_9 --
-\N
-\N
-\N
-\N
-\N
-\N
-\N
-\N
-\N
-\N
+0
 
 -- !gh_data_10 --
-49390617
-64890096
-10696700
-33066637
-32271952
-2051941
-32271952
-57325392
-42386044
-73801003
+19829
 
 -- !gh_data_11 --
-27     
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:27Z","id":"14690746717","public":1,"actor":{"gravatar_id":"","display_login":"sergdudnik","url":"https://api.github.com/users/sergdudnik","id":16341546,"login":"sergdudnik","avatar_url":"https://avatars.githubusercontent.com/u/16341546?"},"repo":{"url":"https://api.github.com/repos/leonardomso/33-js-concepts","id":147350463,"name":"leonardomso/33-js-concepts"},"type":"WatchEvent"}
-36     
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:27Z","id":"14690746732","public":1,"actor":{"gravatar_id":"","display_login":"juliusHuelsmann","url":"https://api.github.com/users/juliusHuelsmann","id":9212314,"login":"juliusHuelsmann","avatar_url":"https://avatars.githubusercontent.com/u/9212314?"},"repo":{"url":"https://api.github.com/repos/odeke-em/drive","id":26109545,"name":"odeke-em/drive"},"type":"WatchEvent"}
-46     
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:27Z","id":"14690746749","org":{"gravatar_id":"","url":"https://api.github.com/orgs/GO-LiFE","id":38434522,"login":"GO-LiFE","avatar_url":"https://avatars.githubusercontent.com/u/38434522?"},"public":1,"actor":{"gravatar_id":"","display_login":"okbean","url":"https://api.github.com/users/okbean","id":75969386,"login":"okbean","avatar_url":"https://avatars.githubusercontent.com/u/75969386?"},"repo":{"url":"https://api.github
 [...]
-56     
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:28Z","id":"14690746773","public":1,"actor":{"gravatar_id":"","display_login":"PWDream","url":"https://api.github.com/users/PWDream","id":4903755,"login":"PWDream","avatar_url":"https://avatars.githubusercontent.com/u/4903755?"},"repo":{"url":"https://api.github.com/repos/MrXujiang/h5-Dooring","id":289417971,"name":"MrXujiang/h5-Dooring"},"type":"WatchEvent"}
-86     
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:29Z","id":"14690746843","public":1,"actor":{"gravatar_id":"","display_login":"Gui-Yom","url":"https://api.github.com/users/Gui-Yom","id":25181283,"login":"Gui-Yom","avatar_url":"https://avatars.githubusercontent.com/u/25181283?"},"repo":{"url":"https://api.github.com/repos/redsaph/cleartext","id":106453399,"name":"redsaph/cleartext"},"type":"WatchEvent"}
-98     
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:29Z","id":"14690746866","org":{"gravatar_id":"","url":"https://api.github.com/orgs/sherlock-project","id":48293496,"login":"sherlock-project","avatar_url":"https://avatars.githubusercontent.com/u/48293496?"},"public":1,"actor":{"gravatar_id":"","display_login":"humaidk2","url":"https://api.github.com/users/humaidk2","id":12982026,"login":"humaidk2","avatar_url":"https://avatars.githubusercontent.com/u/12982026?"},"repo":{";
 [...]
-101    
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:29Z","id":"14690746870","public":1,"actor":{"gravatar_id":"","display_login":"hasantezcan","url":"https://api.github.com/users/hasantezcan","id":32804505,"login":"hasantezcan","avatar_url":"https://avatars.githubusercontent.com/u/32804505?"},"repo":{"url":"https://api.github.com/repos/okandavut/react-spotify-nowplaying","id":326215605,"name":"okandavut/react-spotify-nowplaying"},"type":"WatchEvent"}
-112    
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:30Z","id":"14690746899","public":1,"actor":{"gravatar_id":"","display_login":"nicholas-robertson","url":"https://api.github.com/users/nicholas-robertson","id":17681331,"login":"nicholas-robertson","avatar_url":"https://avatars.githubusercontent.com/u/17681331?"},"repo":{"url":"https://api.github.com/repos/sentriz/gonic","id":178435468,"name":"sentriz/gonic"},"type":"WatchEvent"}
-122    
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:30Z","id":"14690746914","org":{"gravatar_id":"","url":"https://api.github.com/orgs/netlify-labs","id":47546088,"login":"netlify-labs","avatar_url":"https://avatars.githubusercontent.com/u/47546088?"},"public":1,"actor":{"gravatar_id":"","display_login":"javaniecampbell","url":"https://api.github.com/users/javaniecampbell","id":1676496,"login":"javaniecampbell","avatar_url":"https://avatars.githubusercontent.com/u/1676496?
 [...]
-169    
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:32Z","id":"14690747028","org":{"gravatar_id":"","url":"https://api.github.com/orgs/microsoft","id":6154722,"login":"microsoft","avatar_url":"https://avatars.githubusercontent.com/u/6154722?"},"public":1,"actor":{"gravatar_id":"","display_login":"Yxnt","url":"https://api.github.com/users/Yxnt","id":10323352,"login":"Yxnt","avatar_url":"https://avatars.githubusercontent.com/u/10323352?"},"repo":{"url":"https://api.github.co
 [...]
+27     
{"url":"https://api.github.com/repos/leonardomso/33-js-concepts","id":147350463,"name":"leonardomso/33-js-concepts"}
+36     
{"url":"https://api.github.com/repos/odeke-em/drive","id":26109545,"name":"odeke-em/drive"}
+46     
{"url":"https://api.github.com/repos/GO-LiFE/GoFIT_SDK_Android","id":141905736,"name":"GO-LiFE/GoFIT_SDK_Android"}
+56     
{"url":"https://api.github.com/repos/MrXujiang/h5-Dooring","id":289417971,"name":"MrXujiang/h5-Dooring"}
+86     
{"url":"https://api.github.com/repos/redsaph/cleartext","id":106453399,"name":"redsaph/cleartext"}
+98     
{"url":"https://api.github.com/repos/sherlock-project/sherlock","id":162998479,"name":"sherlock-project/sherlock"}
+101    
{"url":"https://api.github.com/repos/okandavut/react-spotify-nowplaying","id":326215605,"name":"okandavut/react-spotify-nowplaying"}
+112    
{"url":"https://api.github.com/repos/sentriz/gonic","id":178435468,"name":"sentriz/gonic"}
+122    
{"url":"https://api.github.com/repos/netlify-labs/react-netlify-identity-widget","id":182606378,"name":"netlify-labs/react-netlify-identity-widget"}
+169    
{"url":"https://api.github.com/repos/microsoft/BotBuilder-Samples","id":68730444,"name":"microsoft/BotBuilder-Samples"}
 
 -- !gh_data_12 --
 2051941        1
@@ -75,3 +57,27 @@ disclose/diodata     2
 64890096       1
 73801003       1
 
+-- !gh_data_13 --
+27     
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:27Z","id":"14690746717","repo":{"url":"https://api.github.com/repos/leonardomso/33-js-concepts","id":147350463,"name":"leonardomso/33-js-concepts"},"actor":{"gravatar_id":"","display_login":"sergdudnik","url":"https://api.github.com/users/sergdudnik","id":16341546,"login":"sergdudnik","avatar_url":"https://avatars.githubusercontent.com/u/16341546?"},"public":1,"type":"WatchEvent"}
  {"payload":{"action":"started"},"created_a [...]
+36     
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:27Z","id":"14690746732","repo":{"url":"https://api.github.com/repos/odeke-em/drive","id":26109545,"name":"odeke-em/drive"},"actor":{"gravatar_id":"","display_login":"juliusHuelsmann","url":"https://api.github.com/users/juliusHuelsmann","id":9212314,"login":"juliusHuelsmann","avatar_url":"https://avatars.githubusercontent.com/u/9212314?"},"public":1,"type":"WatchEvent"}
      {"payload":{"action":"started"},"created_at":"2021-01- [...]
+46     
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:27Z","org":{"gravatar_id":"","url":"https://api.github.com/orgs/GO-LiFE","id":38434522,"login":"GO-LiFE","avatar_url":"https://avatars.githubusercontent.com/u/38434522?"},"id":"14690746749","repo":{"url":"https://api.github.com/repos/GO-LiFE/GoFIT_SDK_Android","id":141905736,"name":"GO-LiFE/GoFIT_SDK_Android"},"actor":{"gravatar_id":"","display_login":"okbean","url":"https://api.github.com/users/okbean","id":75969386,"logi
 [...]
+56     
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:28Z","id":"14690746773","repo":{"url":"https://api.github.com/repos/MrXujiang/h5-Dooring","id":289417971,"name":"MrXujiang/h5-Dooring"},"actor":{"gravatar_id":"","display_login":"PWDream","url":"https://api.github.com/users/PWDream","id":4903755,"login":"PWDream","avatar_url":"https://avatars.githubusercontent.com/u/4903755?"},"public":1,"type":"WatchEvent"}
 {"payload":{"action":"started"},"created_at":"2021-01-02T16:37:28 [...]
+86     
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:29Z","id":"14690746843","repo":{"url":"https://api.github.com/repos/redsaph/cleartext","id":106453399,"name":"redsaph/cleartext"},"actor":{"gravatar_id":"","display_login":"Gui-Yom","url":"https://api.github.com/users/Gui-Yom","id":25181283,"login":"Gui-Yom","avatar_url":"https://avatars.githubusercontent.com/u/25181283?"},"public":1,"type":"WatchEvent"}
     {"payload":{"action":"started"},"created_at":"2021-01-02T16:37:29Z"," [...]
+98     
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:29Z","org":{"gravatar_id":"","url":"https://api.github.com/orgs/sherlock-project","id":48293496,"login":"sherlock-project","avatar_url":"https://avatars.githubusercontent.com/u/48293496?"},"id":"14690746866","repo":{"url":"https://api.github.com/repos/sherlock-project/sherlock","id":162998479,"name":"sherlock-project/sherlock"},"actor":{"gravatar_id":"","display_login":"humaidk2","url":"https://api.github.com/users/humaidk
 [...]
+101    
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:29Z","id":"14690746870","repo":{"url":"https://api.github.com/repos/okandavut/react-spotify-nowplaying","id":326215605,"name":"okandavut/react-spotify-nowplaying"},"actor":{"gravatar_id":"","display_login":"hasantezcan","url":"https://api.github.com/users/hasantezcan","id":32804505,"login":"hasantezcan","avatar_url":"https://avatars.githubusercontent.com/u/32804505?"},"public":1,"type":"WatchEvent"}
       {"payload":{"action":" [...]
+112    
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:30Z","id":"14690746899","repo":{"url":"https://api.github.com/repos/sentriz/gonic","id":178435468,"name":"sentriz/gonic"},"actor":{"gravatar_id":"","display_login":"nicholas-robertson","url":"https://api.github.com/users/nicholas-robertson","id":17681331,"login":"nicholas-robertson","avatar_url":"https://avatars.githubusercontent.com/u/17681331?"},"public":1,"type":"WatchEvent"}
    {"payload":{"action":"started"},"created_at [...]
+122    
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:30Z","org":{"gravatar_id":"","url":"https://api.github.com/orgs/netlify-labs","id":47546088,"login":"netlify-labs","avatar_url":"https://avatars.githubusercontent.com/u/47546088?"},"id":"14690746914","repo":{"url":"https://api.github.com/repos/netlify-labs/react-netlify-identity-widget","id":182606378,"name":"netlify-labs/react-netlify-identity-widget"},"actor":{"gravatar_id":"","display_login":"javaniecampbell","url":"ht
 [...]
+169    
{"payload":{"action":"started"},"created_at":"2021-01-02T16:37:32Z","org":{"gravatar_id":"","url":"https://api.github.com/orgs/microsoft","id":6154722,"login":"microsoft","avatar_url":"https://avatars.githubusercontent.com/u/6154722?"},"id":"14690747028","repo":{"url":"https://api.github.com/repos/microsoft/BotBuilder-Samples","id":68730444,"name":"microsoft/BotBuilder-Samples"},"actor":{"gravatar_id":"","display_login":"Yxnt","url":"https://api.github.com/users/Yxnt","id":10323352,";
 [...]
+
+-- !gh_data_14 --
+{"payload":{"ref_type":"branch","pusher_type":"user","ref":"dependabot/npm_and_yarn/lodash-4.17.19"},"created_at":"2021-01-02T16:37:26Z","id":"14690746680","repo":{"url":"https://api.github.com/repos/megan777/calculator","id":212350500,"name":"megan777/calculator"},"actor":{"gravatar_id":"","display_login":"dependabot","url":"https://api.github.com/users/dependabot[bot]","id":49699333,"login":"dependabot[bot]","avatar_url":"https://avatars.githubusercontent.com/u/49699333?"},"type":"Dele
 [...]
+{"payload":{"ref_type":"branch","pusher_type":"user","ref":"dependabot/npm_and_yarn/acorn-7.4.0"},"created_at":"2021-01-02T16:37:26Z","id":"14690746695","repo":{"url":"https://api.github.com/repos/megan777/calculator","id":212350500,"name":"megan777/calculator"},"actor":{"gravatar_id":"","display_login":"dependabot","url":"https://api.github.com/users/dependabot[bot]","id":49699333,"login":"dependabot[bot]","avatar_url":"https://avatars.githubusercontent.com/u/49699333?"},"type":"DeleteE
 [...]
+{"payload":{"ref_type":"repository","pusher_type":"user","master_branch":"master","description":"Pratique
 Bootstrap 
Openclassroom"},"created_at":"2021-01-02T16:37:27Z","id":"14690746705","repo":{"url":"https://api.github.com/repos/Sonfresh/pratique-bootstrap","id":326225102,"name":"Sonfresh/pratique-bootstrap"},"actor":{"gravatar_id":"","display_login":"Sonfresh","url":"https://api.github.com/users/Sonfresh","id":40009609,"login":"Sonfresh","avatar_url":"https://avatars.githubusercontent
 [...]
+{"payload":{"ref_type":"branch","pusher_type":"user","ref":"dependabot/npm_and_yarn/websocket-extensions-0.1.4"},"created_at":"2021-01-02T16:37:27Z","id":"14690746708","repo":{"url":"https://api.github.com/repos/megan777/calculator","id":212350500,"name":"megan777/calculator"},"actor":{"gravatar_id":"","display_login":"dependabot","url":"https://api.github.com/users/dependabot[bot]","id":49699333,"login":"dependabot[bot]","avatar_url":"https://avatars.githubusercontent.com/u/49699333?"},
 [...]
+{"payload":{"ref_type":"tag","pusher_type":"user","master_branch":"master","description":"A
 library for Frege 
indexers","ref":"v_0.1"},"created_at":"2021-01-02T16:37:27Z","org":{"gravatar_id":"","url":"https://api.github.com/orgs/Software-Engineering-Jagiellonian","id":74735360,"login":"Software-Engineering-Jagiellonian","avatar_url":"https://avatars.githubusercontent.com/u/74735360?"},"id":"14690746727","repo":{"url":"https://api.github.com/repos/Software-Engineering-Jagiellonian/frege-
 [...]
+{"payload":{"ref_type":"repository","pusher_type":"user","master_branch":"master"},"created_at":"2021-01-02T16:37:27Z","id":"14690746737","repo":{"url":"https://api.github.com/repos/TOtto91/orbital-debris","id":326225103,"name":"TOtto91/orbital-debris"},"actor":{"gravatar_id":"","display_login":"TOtto91","url":"https://api.github.com/users/TOtto91","id":54342679,"login":"TOtto91","avatar_url":"https://avatars.githubusercontent.com/u/54342679?"},"type":"CreateEvent","public":1}
+{"payload":{"ref_type":"tag","pusher_type":"user","ref":"flux-write-check-1ccffefc6f"},"created_at":"2021-01-02T16:37:27Z","id":"14690746750","repo":{"url":"https://api.github.com/repos/jimangel/auto-gitops","id":290091833,"name":"jimangel/auto-gitops"},"actor":{"gravatar_id":"","display_login":"jimangel","url":"https://api.github.com/users/jimangel","id":4601051,"login":"jimangel","avatar_url":"https://avatars.githubusercontent.com/u/4601051?"},"type":"DeleteEvent","public":1}
+{"payload":{"ref_type":"branch","pusher_type":"user","master_branch":"master","description":"Open
 Source is hard. Show your 
gratitude","ref":"emoji-conversion"},"created_at":"2021-01-02T16:37:27Z","id":"14690746754","repo":{"url":"https://api.github.com/repos/yeraydiazdiaz/shoutouts.dev","id":323309255,"name":"yeraydiazdiaz/shoutouts.dev"},"actor":{"gravatar_id":"","display_login":"yeraydiazdiaz","url":"https://api.github.com/users/yeraydiazdiaz","id":6739793,"login":"yeraydiazdiaz","ava
 [...]
+{"payload":{"ref_type":"repository","pusher_type":"user","master_branch":"master","description":"BIPN145
 
Winter2020"},"created_at":"2021-01-02T16:37:27Z","id":"14690746759","repo":{"url":"https://api.github.com/repos/mwchu/BIPN145","id":326225104,"name":"mwchu/BIPN145"},"actor":{"gravatar_id":"","display_login":"mwchu","url":"https://api.github.com/users/mwchu","id":76878377,"login":"mwchu","avatar_url":"https://avatars.githubusercontent.com/u/76878377?"},"type":"CreateEvent","public":1}
+{"payload":{"ref_type":"branch","pusher_type":"user","master_branch":"master","description":"Simple
 Python 3 CLI script for retrieving Wrocław MPK trams and buses real-time 
location","ref":"fill-readme"},"created_at":"2021-01-02T16:37:28Z","id":"14690746775","repo":{"url":"https://api.github.com/repos/emkor/mpyk","id":186781673,"name":"emkor/mpyk"},"actor":{"gravatar_id":"","display_login":"emkor","url":"https://api.github.com/users/emkor","id":4871763,"login":"emkor","avatar_url":"https
 [...]
+
diff --git a/regression-test/suites/variant_p0/column_size_limit.groovy 
b/regression-test/suites/variant_p0/column_size_limit.groovy
index 8c8c9051585..70567d89c07 100644
--- a/regression-test/suites/variant_p0/column_size_limit.groovy
+++ b/regression-test/suites/variant_p0/column_size_limit.groovy
@@ -14,8 +14,9 @@
 // KIND, either express or implied.  See the License for the
 // specific language governing permissions and limitations
 // under the License.
+import groovy.json.JsonBuilder
 
-suite("regression_test_variant_column_limit"){
+suite("regression_test_variant_column_limit", "nonConcurrent"){
     def set_be_config = { key, value ->
         String backend_id;
         def backendId_to_backendIP = [:]
@@ -27,7 +28,6 @@ suite("regression_test_variant_column_limit"){
         logger.info("update config: code=" + code + ", out=" + out + ", err=" 
+ err)
     }
     def table_name = "var_column_limit"
-    set_be_config.call("variant_max_merged_tablet_schema_size", "5")
     sql "DROP TABLE IF EXISTS ${table_name}"
     sql """
         CREATE TABLE IF NOT EXISTS ${table_name} (
@@ -39,11 +39,21 @@ suite("regression_test_variant_column_limit"){
         properties("replication_num" = "1", "disable_auto_compaction" = 
"false");
     """
     try {
-        sql """insert into ${table_name} values (1, '{"a" : 1, "b" : 2, "c" : 
3, "d" : 4, "e" : 5}')"""
+        def jsonBuilder = new JsonBuilder()
+        def root = jsonBuilder {
+            // Generate 2049 fields
+            (1..2049).each { fieldNumber ->
+                "field$fieldNumber" fieldNumber
+            }
+        }
+
+        String jsonString = jsonBuilder.toPrettyString()
+        sql """insert into ${table_name} values (1, '$jsonString')"""
     } catch(Exception ex) {
         logger.info("""INSERT INTO ${table_name} failed: """ + ex)
+        assertTrue(ex.toString().contains("Reached max column"));
+    } finally {
     }
     sql """insert into ${table_name} values (1, '{"a" : 1, "b" : 2, "c" : 
3}')"""
 
-    set_be_config.call("variant_max_merged_tablet_schema_size", "1024")
 }
\ No newline at end of file
diff --git a/regression-test/suites/variant_p0/load.groovy 
b/regression-test/suites/variant_p0/load.groovy
index 778da69de96..95d17cff833 100644
--- a/regression-test/suites/variant_p0/load.groovy
+++ b/regression-test/suites/variant_p0/load.groovy
@@ -287,7 +287,7 @@ suite("regression_test_variant", "variant_type"){
         // 12. streamload remote file
         table_name = "logdata"
         create_table.call(table_name, "4")
-        sql "set enable_two_phase_read_opt = false;"
+        // sql "set enable_two_phase_read_opt = false;"
         // no sparse columns
         set_be_config.call("ratio_of_defaults_as_sparse_column", "1")
         load_json_data.call(table_name, """${getS3Url() + 
'/load/logdata.json'}""")
@@ -389,13 +389,13 @@ suite("regression_test_variant", "variant_type"){
         sql """insert into ${table_name} values (2, "abe", '{"c" : 1}')"""
         sql """insert into ${table_name} values (3, "abd", '{"d" : 1}')"""
         sql "delete from ${table_name} where k in (select k from variant_mow 
where k in (1, 2))"
-        qt_sql_38 "select * from ${table_name} order by k"
+        qt_sql_38 "select * from ${table_name} order by k limit 10"
 
         // read text from sparse col
         set_be_config.call("ratio_of_defaults_as_sparse_column", "0")
         sql """insert into  sparse_columns select 0, '{"a": 1123, "b" : [123, 
{"xx" : 1}], "c" : {"c" : 456, "d" : null, "e" : 7.111}, "zzz" : null, "oooo" : 
{"akakaka" : null, "xxxx" : {"xxx" : 123}}}'  as json_str
             union  all select 0, '{"a" : 1234, "xxxx" : "kaana", "ddd" : 
{"aaa" : 123, "mxmxm" : [456, "789"]}}' as json_str from numbers("number" = 
"4096") limit 4096 ;"""
-        qt_sql_31 """select cast(v:xxxx as string) from sparse_columns where 
cast(v:xxxx as string) != 'null' limit 1;"""
+        qt_sql_31 """select cast(v:xxxx as string) from sparse_columns where 
cast(v:xxxx as string) != 'null' order by k limit 1;"""
         sql "truncate table sparse_columns"
         set_be_config.call("ratio_of_defaults_as_sparse_column", "0.95")
 
diff --git a/regression-test/suites/variant_p0/sql/gh_data.sql 
b/regression-test/suites/variant_p0/sql/gh_data.sql
index 7eae73a9faa..41369190ca8 100644
--- a/regression-test/suites/variant_p0/sql/gh_data.sql
+++ b/regression-test/suites/variant_p0/sql/gh_data.sql
@@ -1,4 +1,6 @@
 set exec_mem_limit=8G;
+set enable_two_phase_read_opt = true;
+set global topn_opt_limit_threshold = 1024; 
 SELECT count() from ghdata;
 SELECT cast(v:repo.name as string), count() AS stars FROM ghdata WHERE 
cast(v:type as string) = 'WatchEvent' GROUP BY cast(v:repo.name as string)  
ORDER BY stars DESC, cast(v:repo.name as string) LIMIT 5;
 SELECT max(cast(cast(v:`id` as string) as bigint)) FROM ghdata;
@@ -6,8 +8,9 @@ SELECT sum(cast(cast(v:`id` as string) as bigint)) FROM ghdata;
 SELECT sum(cast(v:payload.member.id as bigint)) FROM ghdata;
 SELECT sum(cast(v:payload.pull_request.milestone.creator.site_admin as 
bigint)) FROM ghdata;
 SELECT sum(length(v:payload.pull_request.base.repo.html_url)) FROM ghdata;
-SELECT v:payload.commits.author.name FROM ghdata ORDER BY k LIMIT 10;
-SELECT v:payload.member.id FROM ghdata where cast(v:payload.member.id as 
string) is not null  ORDER BY k LIMIT 10;
+-- SELECT v:payload.member.id FROM ghdata where cast(v:payload.member.id as 
string) is not null  ORDER BY k LIMIT 10;
 -- select k, v:payload.commits.author.name AS name, e FROM ghdata as t lateral 
view  explode(cast(v:payload.commits.author.name as array<string>)) tm1 as e  
order by k limit 5;
-select k, v from ghdata WHERE cast(v:type as string) = 'WatchEvent'  order by 
k limit 10;
-SELECT cast(v:payload.member.id as bigint), count() FROM ghdata where 
cast(v:payload.member.id as bigint) is not null group by 
cast(v:payload.member.id as bigint) order by 1, 2 desc LIMIT 10;
\ No newline at end of file
+select k, json_extract(v, '$.repo') from ghdata WHERE cast(v:type as string) = 
'WatchEvent'  order by k limit 10;
+SELECT cast(v:payload.member.id as bigint), count() FROM ghdata where 
cast(v:payload.member.id as bigint) is not null group by 
cast(v:payload.member.id as bigint) order by 1, 2 desc LIMIT 10;
+select k, cast(v:`id` as string), cast(v:type as string), cast(v:repo.name as 
string) from ghdata WHERE cast(v:type as string) = 'WatchEvent'  order by k 
limit 10;
+SELECT cast(v:payload.pusher_type as text) FROM ghdata where 
cast(v:payload.pusher_type as text) is not null ORDER BY k LIMIT 10;
\ No newline at end of file


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to