[incubator-doris] branch master updated: [Bug][segment_v2] Fix a bug that NullBitmapBuilder is not reset when data page doesn't have null (#3240)

[zh...@apache.org]

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

zhaoc pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/incubator-doris.git


The following commit(s) were added to refs/heads/master by this push:
     new 8a2eb8f  [Bug][segment_v2] Fix a bug that NullBitmapBuilder is not reset when data page doesn't have null (#3240)
8a2eb8f is described below

commit 8a2eb8fbcf5e168cf2a7459f1a58acf6da563646
Author: Dayue Gao <ga...@meituan.com>
AuthorDate: Wed Apr 1 18:39:04 2020 +0800

    [Bug][segment_v2] Fix a bug that NullBitmapBuilder is not reset when data page doesn't have null (#3240)
    
    This CL fixes a bug that could cause wrong answer for beta rowset with nullable column. The root cause is that NullBitmapBuilder is not reset when the current page doesn't contain NULL, which leads to wrong null map to be written for the next page.
    
    Added a test case to reproduce the problem.
---
 be/src/olap/rowset/segment_v2/column_writer.cpp           |  4 +++-
 .../olap/rowset/segment_v2/column_reader_writer_test.cpp  | 15 +++++++++++++++
 2 files changed, 18 insertions(+), 1 deletion(-)

diff --git a/be/src/olap/rowset/segment_v2/column_writer.cpp b/be/src/olap/rowset/segment_v2/column_writer.cpp
index 62170aa..1ba59bf 100644
--- a/be/src/olap/rowset/segment_v2/column_writer.cpp
+++ b/be/src/olap/rowset/segment_v2/column_writer.cpp
@@ -332,9 +332,11 @@ Status ColumnWriter::_finish_current_page() {
     OwnedSlice nullmap;
     if (_is_nullable && _null_bitmap_builder->has_null()) {
         nullmap = _null_bitmap_builder->finish();
-        _null_bitmap_builder->reset();
         body.push_back(nullmap.slice());
     }
+    if (_null_bitmap_builder != nullptr) {
+        _null_bitmap_builder->reset();
+    }
 
     // prepare data page footer
     std::unique_ptr<Page> page(new Page());
diff --git a/be/test/olap/rowset/segment_v2/column_reader_writer_test.cpp b/be/test/olap/rowset/segment_v2/column_reader_writer_test.cpp
index a67a4f6..c8f6820 100644
--- a/be/test/olap/rowset/segment_v2/column_reader_writer_test.cpp
+++ b/be/test/olap/rowset/segment_v2/column_reader_writer_test.cpp
@@ -313,6 +313,20 @@ TEST_F(ColumnReaderWriterTest, test_nullable) {
     test_nullable_data<OLAP_FIELD_TYPE_BIGINT, BIT_SHUFFLE>(val, is_null, num_uint8_rows / 8, "null_bigint_bs");
     test_nullable_data<OLAP_FIELD_TYPE_LARGEINT, BIT_SHUFFLE>(val, is_null, num_uint8_rows / 16, "null_largeint_bs");
 
+    // test for the case where most values are not null
+    uint8_t* is_null_sparse = new uint8_t[num_uint8_rows];
+    for (int i = 0; i < num_uint8_rows; ++i) {
+        bool v = false;
+        // in order to make some data pages not null, set the first half of values not null.
+        // for the second half, only 1/1024 of values are null
+        if (i >= (num_uint8_rows / 2)) {
+            v = (i % 1024) == 10;
+        }
+        BitmapChange(is_null_sparse, i, v);
+    }
+    test_nullable_data<OLAP_FIELD_TYPE_TINYINT, BIT_SHUFFLE>(val, is_null_sparse, num_uint8_rows, "sparse_null_tiny_bs");
+
+
     float* float_vals = new float[num_uint8_rows];
     for (int i = 0; i < num_uint8_rows; ++i) {
         float_vals[i] = i;
@@ -330,6 +344,7 @@ TEST_F(ColumnReaderWriterTest, test_nullable) {
     // test_nullable_data<OLAP_FIELD_TYPE_DOUBLE, BIT_SHUFFLE>(val, is_null, num_uint8_rows / 8, "null_double_bs");
     delete[] val;
     delete[] is_null;
+    delete[] is_null_sparse;
     delete[] float_vals;
     delete[] double_vals;
 }


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@doris.apache.org
For additional commands, e-mail: commits-help@doris.apache.org