[Fix](flexible partial update) Fix some problems in flexible partial update #41701
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Thank you for your contribution to Apache Doris. Since 2024-03-18, the Document has been moved to doris-website. |
Contributor
Author
|
run buildall |
|
TeamCity be ut coverage result: |
24069a4 to
378ef2f
Compare
d21f1b1 to
d2e1b4c
Compare
6428fa3 to
f177418
Compare
Contributor
Author
|
run buildall |
|
TeamCity be ut coverage result: |
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 40999 ms |
|
TeamCity be ut coverage result: |
TPC-DS: Total hot run time: 192400 ms |
ClickBench: Total hot run time: 32.6 s |
79fe26c to
02cb8af
Compare
Contributor
Author
|
run buildall |
|
TeamCity be ut coverage result: |
TPC-H: Total hot run time: 40756 ms |
TPC-DS: Total hot run time: 192006 ms |
ClickBench: Total hot run time: 33.62 s |
dataroaring
reviewed
Oct 14, 2024
dataroaring
reviewed
Oct 14, 2024
Contributor
Author
|
run buildall |
Contributor
Cloud UT Coverage ReportIncrement line coverage Increment coverage report
|
TPC-H: Total hot run time: 33716 ms |
TPC-DS: Total hot run time: 186267 ms |
ClickBench: Total hot run time: 30.31 s |
Contributor
BE UT Coverage ReportIncrement line coverage Increment coverage report
|
Contributor
Author
|
run cloud_p0 |
Contributor
|
PR approved by at least one committer and no changes requested. |
Contributor
|
PR approved by anyone and no changes requested. |
bobhan1
added a commit
to bobhan1/doris
that referenced
this pull request
Jul 18, 2025
…update (apache#41701) 1. Fix incorrect result when there are insert after delete for same key in one load. In flexible partial update, if there is an insert after a delete on the same key in one batch, the final result will be always delete that row in table, which is incorrect. The reason is that when doing aggregation for rows with same keys in memtable, we treat delete sign column as an ordinary column. So if there is a row with delete sign in that batch, the final aggregated row will always have a delete sign and any insert on the same key after it will be lost. In the above situation, we should keep those rows and don't merge them when doing aggregation in memtable. When flush the memtable to segment, we should first apply the semantic of delete if exists and then insert the row after it. 2. fix flexible partial update don't use row store column to read data 3. Fix flexible partial update incorrect read index when resolving conflict in publish phase 4. add some default value cases and fix on update current timestamp 5. fix publish phase alignment problem when table has sequence column and the rowset to read has multi segments. 6. add some publish alignment case when table has sequence column 7. ~revert apache#25346, we use the sequence value in flush phase as its final value and will not do alignment in publish phase if the partial update load doesn't specified sequence column.~ 8. fix auto-increment column value is wrongly updated when it's not specified. 9. Fix the problem that the value of sequence column may decrease. 10. Only read seq col and seq map col when the previous row is deleted in partial update and flexible partial update when doing alignment. 11. Fix the value of auto-increment column is not consistent between replicas.
bobhan1
added a commit
to bobhan1/doris
that referenced
this pull request
Jul 18, 2025
…update (apache#41701) 1. Fix incorrect result when there are insert after delete for same key in one load. In flexible partial update, if there is an insert after a delete on the same key in one batch, the final result will be always delete that row in table, which is incorrect. The reason is that when doing aggregation for rows with same keys in memtable, we treat delete sign column as an ordinary column. So if there is a row with delete sign in that batch, the final aggregated row will always have a delete sign and any insert on the same key after it will be lost. In the above situation, we should keep those rows and don't merge them when doing aggregation in memtable. When flush the memtable to segment, we should first apply the semantic of delete if exists and then insert the row after it. 2. fix flexible partial update don't use row store column to read data 3. Fix flexible partial update incorrect read index when resolving conflict in publish phase 4. add some default value cases and fix on update current timestamp 5. fix publish phase alignment problem when table has sequence column and the rowset to read has multi segments. 6. add some publish alignment case when table has sequence column 7. ~revert apache#25346, we use the sequence value in flush phase as its final value and will not do alignment in publish phase if the partial update load doesn't specified sequence column.~ 8. fix auto-increment column value is wrongly updated when it's not specified. 9. Fix the problem that the value of sequence column may decrease. 10. Only read seq col and seq map col when the previous row is deleted in partial update and flexible partial update when doing alignment. 11. Fix the value of auto-increment column is not consistent between replicas.
bobhan1
added a commit
to bobhan1/doris
that referenced
this pull request
Jul 21, 2025
…update (apache#41701) 1. Fix incorrect result when there are insert after delete for same key in one load. In flexible partial update, if there is an insert after a delete on the same key in one batch, the final result will be always delete that row in table, which is incorrect. The reason is that when doing aggregation for rows with same keys in memtable, we treat delete sign column as an ordinary column. So if there is a row with delete sign in that batch, the final aggregated row will always have a delete sign and any insert on the same key after it will be lost. In the above situation, we should keep those rows and don't merge them when doing aggregation in memtable. When flush the memtable to segment, we should first apply the semantic of delete if exists and then insert the row after it. 2. fix flexible partial update don't use row store column to read data 3. Fix flexible partial update incorrect read index when resolving conflict in publish phase 4. add some default value cases and fix on update current timestamp 5. fix publish phase alignment problem when table has sequence column and the rowset to read has multi segments. 6. add some publish alignment case when table has sequence column 7. ~revert apache#25346, we use the sequence value in flush phase as its final value and will not do alignment in publish phase if the partial update load doesn't specified sequence column.~ 8. fix auto-increment column value is wrongly updated when it's not specified. 9. Fix the problem that the value of sequence column may decrease. 10. Only read seq col and seq map col when the previous row is deleted in partial update and flexible partial update when doing alignment. 11. Fix the value of auto-increment column is not consistent between replicas.
bobhan1
added a commit
to bobhan1/doris
that referenced
this pull request
Jul 22, 2025
…update (apache#41701) 1. Fix incorrect result when there are insert after delete for same key in one load. In flexible partial update, if there is an insert after a delete on the same key in one batch, the final result will be always delete that row in table, which is incorrect. The reason is that when doing aggregation for rows with same keys in memtable, we treat delete sign column as an ordinary column. So if there is a row with delete sign in that batch, the final aggregated row will always have a delete sign and any insert on the same key after it will be lost. In the above situation, we should keep those rows and don't merge them when doing aggregation in memtable. When flush the memtable to segment, we should first apply the semantic of delete if exists and then insert the row after it. 2. fix flexible partial update don't use row store column to read data 3. Fix flexible partial update incorrect read index when resolving conflict in publish phase 4. add some default value cases and fix on update current timestamp 5. fix publish phase alignment problem when table has sequence column and the rowset to read has multi segments. 6. add some publish alignment case when table has sequence column 7. ~revert apache#25346, we use the sequence value in flush phase as its final value and will not do alignment in publish phase if the partial update load doesn't specified sequence column.~ 8. fix auto-increment column value is wrongly updated when it's not specified. 9. Fix the problem that the value of sequence column may decrease. 10. Only read seq col and seq map col when the previous row is deleted in partial update and flexible partial update when doing alignment. 11. Fix the value of auto-increment column is not consistent between replicas.
bobhan1
added a commit
to bobhan1/doris
that referenced
this pull request
Jul 23, 2025
…update (apache#41701) 1. Fix incorrect result when there are insert after delete for same key in one load. In flexible partial update, if there is an insert after a delete on the same key in one batch, the final result will be always delete that row in table, which is incorrect. The reason is that when doing aggregation for rows with same keys in memtable, we treat delete sign column as an ordinary column. So if there is a row with delete sign in that batch, the final aggregated row will always have a delete sign and any insert on the same key after it will be lost. In the above situation, we should keep those rows and don't merge them when doing aggregation in memtable. When flush the memtable to segment, we should first apply the semantic of delete if exists and then insert the row after it. 2. fix flexible partial update don't use row store column to read data 3. Fix flexible partial update incorrect read index when resolving conflict in publish phase 4. add some default value cases and fix on update current timestamp 5. fix publish phase alignment problem when table has sequence column and the rowset to read has multi segments. 6. add some publish alignment case when table has sequence column 7. ~revert apache#25346, we use the sequence value in flush phase as its final value and will not do alignment in publish phase if the partial update load doesn't specified sequence column.~ 8. fix auto-increment column value is wrongly updated when it's not specified. 9. Fix the problem that the value of sequence column may decrease. 10. Only read seq col and seq map col when the previous row is deleted in partial update and flexible partial update when doing alignment. 11. Fix the value of auto-increment column is not consistent between replicas.
16 tasks
zhannngchen
pushed a commit
that referenced
this pull request
Jul 24, 2025
…_update` (#53765) ### What problem does this PR solve? introduced in #41701 When doing alignment for partial update in publish phase, `read_index_old` may be not continuous if some rows in current block has delete signs. Thus `old_block_delete_signs[read_index_old[idx]]` may overflow when all rows in current block have delete signs because `old_block_delete_signs` is empty.
morrySnow
pushed a commit
that referenced
this pull request
Jul 25, 2025
w41ter
pushed a commit
to w41ter/incubator-doris
that referenced
this pull request
Jul 30, 2025
…_update` (apache#53765) ### What problem does this PR solve? introduced in apache#41701 When doing alignment for partial update in publish phase, `read_index_old` may be not continuous if some rows in current block has delete signs. Thus `old_block_delete_signs[read_index_old[idx]]` may overflow when all rows in current block have delete signs because `old_block_delete_signs` is empty.
zhiqiang-hhhh
pushed a commit
to zhiqiang-hhhh/doris
that referenced
this pull request
Aug 4, 2025
…_update` (apache#53765) ### What problem does this PR solve? introduced in apache#41701 When doing alignment for partial update in publish phase, `read_index_old` may be not continuous if some rows in current block has delete signs. Thus `old_block_delete_signs[read_index_old[idx]]` may overflow when all rows in current block have delete signs because `old_block_delete_signs` is empty.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In the above situation, we should keep those rows and don't merge them when doing aggregation in memtable. When flush the memtable to segment, we should first apply the semantic of delete if exists and then insert the row after it.
revert [Fix](partial-update) Correct the alignment process when the table has sequence column and add cases #25346, we use the sequence value in flush phase as its final value and will not do alignment in publish phase if the partial update load doesn't specified sequence column.