[doris] branch master updated: [bugfix](predicate column) data maybe wrong if not a single page (#12796)

[yiguolei]

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

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


The following commit(s) were added to refs/heads/master by this push:
     new 32551a7263 [bugfix](predicate column) data maybe wrong if not a single 
page (#12796)
32551a7263 is described below

commit 32551a7263369843214a7da02bebe43756183b56
Author: yiguolei <676222...@qq.com>
AuthorDate: Thu Sep 22 09:55:31 2022 +0800

    [bugfix](predicate column) data maybe wrong if not a single page (#12796)
    
    
    Co-authored-by: yiguolei <yiguo...@gmail.com>
---
 be/src/vec/columns/column_string.h    |  5 ++---
 be/src/vec/columns/predicate_column.h | 22 ++++++++++++++++++----
 2 files changed, 20 insertions(+), 7 deletions(-)

diff --git a/be/src/vec/columns/column_string.h 
b/be/src/vec/columns/column_string.h
index 783d79cf07..827ec896a7 100644
--- a/be/src/vec/columns/column_string.h
+++ b/be/src/vec/columns/column_string.h
@@ -168,9 +168,8 @@ public:
         for (size_t i = 0; i < num; i++) {
             uint32_t len = len_array[i];
             uint32_t start_offset = start_offset_array[i];
-            if (len) {
-                memcpy(data + offset, data_array + start_offset, len);
-            }
+            // memcpy will deal len == 0, not do it here
+            memcpy(data + offset, data_array + start_offset, len);
             offset += len;
             offsets.push_back(offset);
         }
diff --git a/be/src/vec/columns/predicate_column.h 
b/be/src/vec/columns/predicate_column.h
index 68a4095450..fd99d4c04b 100644
--- a/be/src/vec/columns/predicate_column.h
+++ b/be/src/vec/columns/predicate_column.h
@@ -278,15 +278,29 @@ public:
             }
 
             char* destination = (char*)_pool->allocate(total_mem_size);
-            memcpy(destination, data_array, total_mem_size);
-            // Resize the underline data to allow data copy directly
+            char* org_dst = destination;
             size_t org_elem_num = data.size();
             data.resize(org_elem_num + num);
+            uint32_t fragment_start_offset = start_offset_array[0];
+            size_t fragment_len = 0;
             for (size_t i = 0; i < num; i++) {
-                data[org_elem_num + i].ptr = destination;
+                data[org_elem_num + i].ptr = destination + fragment_len;
                 data[org_elem_num + i].len = len_array[i];
-                destination += len_array[i];
+                fragment_len += len_array[i];
+                // Compute the largest continuous memcpy block and copy them.
+                // If this is the last element in data array, then should copy 
the current memory block.
+                if (i == num - 1 ||
+                    start_offset_array[i + 1] != start_offset_array[i] + 
len_array[i]) {
+                    memcpy(destination, data_array + fragment_start_offset, 
fragment_len);
+                    destination += fragment_len;
+                    fragment_start_offset = (i == num - 1 ? 0 : 
start_offset_array[i + 1]);
+                    fragment_len = 0;
+                }
             }
+            CHECK(destination - org_dst == total_mem_size)
+                    << "Copied size not equal to expected size";
+        } else {
+            LOG(FATAL) << "Method insert_many_binary_data is not supported";
         }
     }
 


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