[GitHub] carbondata pull request #2574: [HotFix][CARBONDATA-2788][BloomDataMap] Fix b...

[qiuchenjian-2]

GitHub user xuchuanyin opened a pull request:

    https://github.com/apache/carbondata/pull/2574

    [HotFix][CARBONDATA-2788][BloomDataMap] Fix bugs in incorrect query result with bloom datamap

    This PR solve three problems which will affect the correctness of the query on bloom.
   
    1. optimize blockletId in rebuilding datamap
    After review the code, we found that modification in PR2539 is not needed, so we revert that PR.
   
    2. bugs in overflow for blocklet count
    Carbondata stores blocklet count for each block in byte data type, when
    a block contains more than 128 blocklets, it will overflow the byte
    limits.
    Here we change the data type to short.
   
    3. Fix bug in querying with bloom datamap with block cache level enabled
    In block cache level scenario, previously the main BlockDataMap return
    block as pruned blocklet with its blockletId=-1;
    However, other index datamap such as BloomDataMap return actual blocklet
    with correct blockletId. Due to the behaviour of Blocklet's hashcode,
    some blocklets will be uncorrectly marked as duplicated and dropped.
    Thus cause incorrect query result.
   
    To fix this problem, we will return all blocklets with correct
    blockletId for the block instead of returning a fake blocklet with
    blockletId=-1.
    This will not affect the following procedure.
   
    Be sure to do all of the following checklist to help us incorporate
    your contribution quickly and easily:
   
     - [ ] Any interfaces changed?
     
     - [ ] Any backward compatibility impacted?
     
     - [ ] Document update required?
   
     - [ ] Testing done
            Please provide details on
            - Whether new unit test cases have been added or why no new tests are required?
            - How it is tested? Please attach test report.
            - Is it a performance related change? Please attach the performance test report.
            - Any additional information to help reviewers in testing this change.
           
     - [ ] For large changes, please consider breaking it into sub-tasks under an umbrella JIRA.
   


You can merge this pull request into a Git repository by running:

    $ git pull https://github.com/xuchuanyin/carbondata 0728_fix_bug_query_bloom_opt

Alternatively you can review and apply these changes as the patch at:

    https://github.com/apache/carbondata/pull/2574.patch

To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:

    This closes #2574
   
----
commit cba17975affad343d555f49cb255043b556310f7
Author: xuchuanyin <xuchuanyin@...>
Date:   2018-07-27T13:13:51Z

    Fix bugs in overflow for blocklet count
   
    Carbondata stores blocklet count for each block in byte data type, when
    a block contains more than 128 blocklets, it will overflow the byte
    limits.
    Here we change the data type to short.

commit 7e41d9533dae86ab26e102a0342c128cbab03d1c
Author: xuchuanyin <xuchuanyin@...>
Date:   2018-07-28T07:37:38Z

    Revert optimize blockletId in rebuilding datamap
   
    We found querying huge data with rebuilding bloom datamap will give
    incorrect result. The root cause is that the blockletId in
    ResultCollector is wrong. (This was introduced in PR2539)
    We will revert the previous modification for this. Now it is checked and
    works fine.

commit e418018e05cb0a15c996c5bb58debb0486252f84
Author: xuchuanyin <xuchuanyin@...>
Date:   2018-07-28T09:08:52Z

    Fix bug in querying with bloom datamap with block cache level enabled
   
    In block cache level scenario, previously the main BlockDataMap return
    block as pruned blocklet with its blockletId=-1;
    However, other index datamap such as BloomDataMap return actual blocklet
    with correct blockletId. Due to the behaviour of Blocklet's hashcode,
    some blocklets will be uncorrectly marked as duplicated and dropped.
    Thus cause incorrect query result.
   
    To fix this problem, we will return all blocklets with correct
    blockletId for the block instead of returning a fake blocklet with
    blockletId=-1.
    This will not affect the following procedure.

----


---

[qiuchenjian-2]

Github user xuchuanyin commented on the issue:

    https://github.com/apache/carbondata/pull/2574
 
    hi @manishgupta88, can you check this PR. It is intended to replace PR #2565


---

[qiuchenjian-2]

Github user xuchuanyin commented on the issue:

    https://github.com/apache/carbondata/pull/2574
 
    hi @ravipesala can you check this PR also.
    It contains three problems, you can refer to the description of the PR for more information


---

[qiuchenjian-2]

Github user CarbonDataQA commented on the issue:

    https://github.com/apache/carbondata/pull/2574
 
    Build Success with Spark 2.1.0, Please check CI http://136.243.101.176:8080/job/ApacheCarbonPRBuilder1/7582/



---

[qiuchenjian-2]

Github user CarbonDataQA commented on the issue:

    https://github.com/apache/carbondata/pull/2574
 
    Build Failed with Spark 2.2.1, Please check CI http://88.99.58.216:8080/job/ApacheCarbonPRBuilder/6337/



---

[qiuchenjian-2]

Github user ravipesala commented on the issue:

    https://github.com/apache/carbondata/pull/2574
 
    LGTM , @manishgupta88 Please review once


---

[qiuchenjian-2]

Github user xuchuanyin commented on the issue:

    https://github.com/apache/carbondata/pull/2574
 
    refactored the pruning procedure for BlockDataMap


---

[qiuchenjian-2]

Github user ravipesala commented on the issue:

    https://github.com/apache/carbondata/pull/2574
 
    SDV Build Fail , Please check CI http://144.76.159.231:8080/job/ApacheSDVTests/6043/



---

[qiuchenjian-2]

Github user CarbonDataQA commented on the issue:

    https://github.com/apache/carbondata/pull/2574
 
    Build Failed  with Spark 2.1.0, Please check CI http://136.243.101.176:8080/job/ApacheCarbonPRBuilder1/7591/



---

[qiuchenjian-2]

Github user manishgupta88 commented on a diff in the pull request:

    https://github.com/apache/carbondata/pull/2574#discussion_r205948964
 
    --- Diff: core/src/main/java/org/apache/carbondata/core/indexstore/blockletindex/BlockletDataMap.java ---
    @@ -223,17 +222,15 @@ protected short getBlockletId(DataMapRow dataMapRow) {
         return dataMapRow.getShort(BLOCKLET_ID_INDEX);
       }
     
    -  protected ExtendedBlocklet createBlocklet(DataMapRow row, String fileName, short blockletId,
    -      boolean useMinMaxForPruning) {
    +  protected ExtendedBlocklet createBlocklet(DataMapRow row, String fileName, short blockletId) {
         if (isLegacyStore) {
    -      return super.createBlocklet(row, fileName, blockletId, useMinMaxForPruning);
    +      return super.createBlocklet(row, fileName, blockletId);
    --- End diff --
   
    useMinMaxForPruning flag is added to handle a different scenario where cache_level = blocklet and the filter column min max in not cached in driver. In that case the blocklet pruning for min/max needs to be done in each executor which is identified using this flag.


---

[qiuchenjian-2]

Github user manishgupta88 commented on a diff in the pull request:

    https://github.com/apache/carbondata/pull/2574#discussion_r205948912
 
    --- Diff: core/src/main/java/org/apache/carbondata/core/indexstore/blockletindex/BlockDataMap.java ---
    @@ -542,48 +551,48 @@ protected void createSummaryDMStore(BlockletDataMapModel blockletDataMapModel)
         List<Blocklet> blocklets = new ArrayList<>();
         CarbonRowSchema[] schema = getFileFooterEntrySchema();
         String filePath = getFilePath();
    -    int numBlocklets = 0;
    -    if (filterExp == null) {
    -      numBlocklets = memoryDMStore.getRowCount();
    -      for (int i = 0; i < numBlocklets; i++) {
    -        DataMapRow safeRow = memoryDMStore.getDataMapRow(schema, i).convertToSafeRow();
    -        blocklets.add(createBlocklet(safeRow, getFileNameWithFilePath(safeRow, filePath),
    -            getBlockletId(safeRow), false));
    -      }
    -    } else {
    -      // Remove B-tree jump logic as start and end key prepared is not
    -      // correct for old store scenarios
    -      int startIndex = 0;
    -      numBlocklets = memoryDMStore.getRowCount();
    -      FilterExecuter filterExecuter = FilterUtil
    -          .getFilterExecuterTree(filterExp, getSegmentProperties(), null, getMinMaxCacheColumns());
    -      // flag to be used for deciding whether use min/max in executor pruning for BlockletDataMap
    -      boolean useMinMaxForPruning = useMinMaxForExecutorPruning(filterExp);
    -      // min and max for executor pruning
    -      while (startIndex < numBlocklets) {
    -        DataMapRow safeRow = memoryDMStore.getDataMapRow(schema, startIndex).convertToSafeRow();
    -        String fileName = getFileNameWithFilePath(safeRow, filePath);
    -        short blockletId = getBlockletId(safeRow);
    -        boolean isValid =
    -            addBlockBasedOnMinMaxValue(filterExecuter, getMinMaxValue(safeRow, MAX_VALUES_INDEX),
    -                getMinMaxValue(safeRow, MIN_VALUES_INDEX), fileName, blockletId);
    -        if (isValid) {
    -          blocklets.add(createBlocklet(safeRow, fileName, blockletId, useMinMaxForPruning));
    +    ByteBuffer byteBuffer = ByteBuffer.wrap(getBlockletRowCountForEachBlock());
    --- End diff --
   
    In case of legacy store (store in version < = 1.1), we are not storing the blocklet count as the index file footer does not contain the blocklet information...so this code will throw exception


---

[qiuchenjian-2]

Github user manishgupta88 commented on the issue:

    https://github.com/apache/carbondata/pull/2574
 
    @xuchuanyin ...It is better that we should go with PR #2565  as it contains the proper code which handles the bugs.


---

[qiuchenjian-2]

Github user CarbonDataQA commented on the issue:

    https://github.com/apache/carbondata/pull/2574
 
    Build Failed with Spark 2.2.1, Please check CI http://88.99.58.216:8080/job/ApacheCarbonPRBuilder/6346/



---

[qiuchenjian-2]

Github user ravipesala commented on the issue:

    https://github.com/apache/carbondata/pull/2574
 
    SDV Build Fail , Please check CI http://144.76.159.231:8080/job/ApacheSDVTests/6053/



---

[qiuchenjian-2]

Github user xuchuanyin commented on a diff in the pull request:

    https://github.com/apache/carbondata/pull/2574#discussion_r205995542
 
    --- Diff: core/src/main/java/org/apache/carbondata/core/indexstore/blockletindex/BlockletDataMap.java ---
    @@ -223,17 +222,15 @@ protected short getBlockletId(DataMapRow dataMapRow) {
         return dataMapRow.getShort(BLOCKLET_ID_INDEX);
       }
     
    -  protected ExtendedBlocklet createBlocklet(DataMapRow row, String fileName, short blockletId,
    -      boolean useMinMaxForPruning) {
    +  protected ExtendedBlocklet createBlocklet(DataMapRow row, String fileName, short blockletId) {
         if (isLegacyStore) {
    -      return super.createBlocklet(row, fileName, blockletId, useMinMaxForPruning);
    +      return super.createBlocklet(row, fileName, blockletId);
    --- End diff --
   
    yes, there are test errors about this problem


---

[qiuchenjian-2]

Github user xuchuanyin commented on a diff in the pull request:

    https://github.com/apache/carbondata/pull/2574#discussion_r205996478
 
    --- Diff: core/src/main/java/org/apache/carbondata/core/indexstore/blockletindex/BlockDataMap.java ---
    @@ -542,48 +551,48 @@ protected void createSummaryDMStore(BlockletDataMapModel blockletDataMapModel)
         List<Blocklet> blocklets = new ArrayList<>();
         CarbonRowSchema[] schema = getFileFooterEntrySchema();
         String filePath = getFilePath();
    -    int numBlocklets = 0;
    -    if (filterExp == null) {
    -      numBlocklets = memoryDMStore.getRowCount();
    -      for (int i = 0; i < numBlocklets; i++) {
    -        DataMapRow safeRow = memoryDMStore.getDataMapRow(schema, i).convertToSafeRow();
    -        blocklets.add(createBlocklet(safeRow, getFileNameWithFilePath(safeRow, filePath),
    -            getBlockletId(safeRow), false));
    -      }
    -    } else {
    -      // Remove B-tree jump logic as start and end key prepared is not
    -      // correct for old store scenarios
    -      int startIndex = 0;
    -      numBlocklets = memoryDMStore.getRowCount();
    -      FilterExecuter filterExecuter = FilterUtil
    -          .getFilterExecuterTree(filterExp, getSegmentProperties(), null, getMinMaxCacheColumns());
    -      // flag to be used for deciding whether use min/max in executor pruning for BlockletDataMap
    -      boolean useMinMaxForPruning = useMinMaxForExecutorPruning(filterExp);
    -      // min and max for executor pruning
    -      while (startIndex < numBlocklets) {
    -        DataMapRow safeRow = memoryDMStore.getDataMapRow(schema, startIndex).convertToSafeRow();
    -        String fileName = getFileNameWithFilePath(safeRow, filePath);
    -        short blockletId = getBlockletId(safeRow);
    -        boolean isValid =
    -            addBlockBasedOnMinMaxValue(filterExecuter, getMinMaxValue(safeRow, MAX_VALUES_INDEX),
    -                getMinMaxValue(safeRow, MIN_VALUES_INDEX), fileName, blockletId);
    -        if (isValid) {
    -          blocklets.add(createBlocklet(safeRow, fileName, blockletId, useMinMaxForPruning));
    +    ByteBuffer byteBuffer = ByteBuffer.wrap(getBlockletRowCountForEachBlock());
    --- End diff --
   
    for legacy store, the pruned result of other index datamap will be processed by ‘createBlockletFromRelativeBlockletId’ which will also try to get blocklet count for each block --- will this also be a problem?


---

[qiuchenjian-2]

Github user xuchuanyin commented on the issue:

    https://github.com/apache/carbondata/pull/2574
 
    @ravipesala @manishgupta88 As we discussed online, we use PR #2565 and update it based on your advices. So I'll close this PR.


---

[qiuchenjian-2]

Github user xuchuanyin closed the pull request at:

    https://github.com/apache/carbondata/pull/2574


---