4paradigm / 4paradigm/OpenMLDB

Wrong result for MemTableTraverseIterator::GetCount()

Open
#1,227 1 comment 0 reactions 1 assignee Assigned to @dl239 View on GitHub
bug storage-engine
Dominant language
C++
Stars
1.7k
Forks
331
Avg merge
12d 12h
Merged PRs (30d)
1

Description

When using traverseIterator to traverse Memtable, GetCount() is used to get the traverse count. The meaning of 'traverse count' is not clear. and if we regard it as 'how many pieces of data it has traversed' then the result is not correct.

## Situation
``` cpp
TEST_F(TableTest, TraverseIteratorCount) {
::openmldb::api::TableMeta table_meta;
table_meta.set_name("table1");
table_meta.set_tid(1);
table_meta.set_pid(0);
table_meta.set_seg_cnt(8);
table_meta.set_mode(::openmldb::api::TableMode::kTableLeader);
table_meta.set_key_entry_max_height(8);
table_meta.set_format_version(1);
SchemaCodec::SetColumnDesc(table_meta.add_column_desc(), "card", ::openmldb::type::kString);
SchemaCodec::SetColumnDesc(table_meta.add_column_desc(), "mcc", ::openmldb::type::kString);
SchemaCodec::SetColumnDesc(table_meta.add_column_desc(), "price", ::openmldb::type::kBigInt);
SchemaCodec::SetColumnDesc(table_meta.add_column_desc(), "ts1", ::openmldb::type::kBigInt);
SchemaCodec::SetColumnDesc(table_meta.add_column_desc(), "ts2", ::openmldb::type::kBigInt);
SchemaCodec::SetIndex(table_meta.add_column_key(), "card", "card", "ts1", ::openmldb::type::kAbsoluteTime, 0, 0);
SchemaCodec::SetIndex(table_meta.add_column_key(), "card1", "card", "ts2", ::openmldb::type::kAbsoluteTime, 0, 0);
SchemaCodec::SetIndex(table_meta.add_column_key(), "mcc", "mcc", "ts1", ::openmldb::type::kAbsoluteTime, 0, 0);

MemTable table(table_meta);
table.Init();
codec::SDKCodec codec(table_meta);

for (int i = 0; i < 1000; i++) {
std::vector row = {"card" + std::to_string(i % 100), "mcc" + std::to_string(i),
"13", std::to_string(1000 + i), std::to_string(10000 + i)};
::openmldb::api::PutRequest request;
::openmldb::api::Dimension* dim = request.add_dimensions();
dim->set_idx(0);
dim->set_key(row[0]);
dim = request.add_dimensions();
dim->set_idx(1);
dim->set_key(row[0]);
dim = request.add_dimensions();
dim->set_idx(2);
dim->set_key(row[1]);
std::string value;
ASSERT_EQ(0, codec.EncodeRow(row, &value));
table.Put(0, value, request.dimensions());
}
TableIterator* it = table.NewTraverseIterator(0);
it->SeekToFirst();
int count = 0;
while (it->Valid()) {
count++;
it->Next();
}
ASSERT_EQ(1000, count);
ASSERT_EQ(1100, (int64_t)it->GetCount());
delete it;
}
```

## Expected Behavior

The result of (int64_t)it->GetCount() should probably be 1000

## Current Behavior

It's 1100 as the test says.

## Steps to Reproduce
make and run table_test, in TEST_F(TableTest, TraverseIteratorCount)

Contributor guide

Open the contributing guide

Research direction

The issue is in MemTableTraverseIterator::GetCount() within the OpenMLDB codebase. Start by examining the table_test.cpp file and the specific test case 'TraverseIteratorCount'. Look at the MemTable and MemTableTraverseIterator implementations to understand how the count is incremented. The test inserts 1000 rows but expects GetCount() to return 1000, not 1100. Run the test to reproduce the failure, then trace the logic in GetCount() to identify why it overcounts.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
databases
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Clearly specified
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.