Return-Path: <dev-return-25145-archive-asf-public=cust-asf.ponee.io@drill.apache.org>
X-Original-To: archive-asf-public-internal@cust-asf2.ponee.io
Delivered-To: archive-asf-public-internal@cust-asf2.ponee.io
Received: from cust-asf.ponee.io (cust-asf.ponee.io [163.172.22.183])
	by cust-asf2.ponee.io (Postfix) with ESMTP id E8B9E200BB6
	for <archive-asf-public-internal@cust-asf2.ponee.io>; Fri,  4 Nov 2016 23:12:48 +0100 (CET)
Received: by cust-asf.ponee.io (Postfix)
	id E7679160B07; Fri,  4 Nov 2016 22:12:48 +0000 (UTC)
Delivered-To: archive-asf-public@cust-asf.ponee.io
Received: from mail.apache.org (hermes.apache.org [140.211.11.3])
	by cust-asf.ponee.io (Postfix) with SMTP id 1666E160B04
	for <archive-asf-public@cust-asf.ponee.io>; Fri,  4 Nov 2016 23:12:47 +0100 (CET)
Received: (qmail 27713 invoked by uid 500); 4 Nov 2016 22:12:46 -0000
Mailing-List: contact dev-help@drill.apache.org; run by ezmlm
Precedence: bulk
List-Help: <mailto:dev-help@drill.apache.org>
List-Unsubscribe: <mailto:dev-unsubscribe@drill.apache.org>
List-Post: <mailto:dev@drill.apache.org>
List-Id: <dev.drill.apache.org>
Reply-To: dev@drill.apache.org
Delivered-To: mailing list dev@drill.apache.org
Received: (qmail 27494 invoked by uid 99); 4 Nov 2016 22:12:46 -0000
Received: from git1-us-west.apache.org (HELO git1-us-west.apache.org) (140.211.11.23)
    by apache.org (qpsmtpd/0.29) with ESMTP; Fri, 04 Nov 2016 22:12:46 +0000
Received: by git1-us-west.apache.org (ASF Mail Server at git1-us-west.apache.org, from userid 33)
	id 489A2E3813; Fri,  4 Nov 2016 22:12:46 +0000 (UTC)
From: parthchandra <git@git.apache.org>
To: dev@drill.apache.org
Reply-To: dev@drill.apache.org
References: <git-pr-611-drill@git.apache.org>
In-Reply-To: <git-pr-611-drill@git.apache.org>
Subject: [GitHub] drill pull request #611: Drill-4800: Improve parquet reader performance
Content-Type: text/plain
Message-Id: <20161104221246.489A2E3813@git1-us-west.apache.org>
Date: Fri,  4 Nov 2016 22:12:46 +0000 (UTC)
archived-at: Fri, 04 Nov 2016 22:12:49 -0000

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

    https://github.com/apache/drill/pull/611#discussion_r86627781
  
    --- Diff: exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/columnreaders/VarLenBinaryReader.java ---
    @@ -43,43 +51,153 @@ public VarLenBinaryReader(ParquetRecordReader parentReader, List<VarLengthColumn
       public long readFields(long recordsToReadInThisPass, ColumnReader<?> firstColumnStatus) throws IOException {
     
         long recordsReadInCurrentPass = 0;
    -    int lengthVarFieldsInCurrentRecord;
    -    long totalVariableLengthData = 0;
    -    boolean exitLengthDeterminingLoop = false;
    +
         // write the first 0 offset
         for (VarLengthColumn<?> columnReader : columns) {
           columnReader.reset();
         }
     
    +    //if(useAsyncTasks){
    +    //  recordsReadInCurrentPass = determineSizesParallel(recordsToReadInThisPass);
    +    //} else {
    +      recordsReadInCurrentPass = determineSizesSerial(recordsToReadInThisPass);
    +    //}
    +    if(useAsyncTasks){
    +      readRecordsParallel(recordsReadInCurrentPass);
    +    }else{
    +      readRecordsSerial(recordsReadInCurrentPass);
    +    }
    +    return recordsReadInCurrentPass;
    +  }
    +
    +
    +  private long determineSizesSerial(long recordsToReadInThisPass) throws IOException {
    +    int lengthVarFieldsInCurrentRecord = 0;
    +    boolean exitLengthDeterminingLoop = false;
    +    long totalVariableLengthData = 0;
    +    long recordsReadInCurrentPass = 0;
         do {
    -      lengthVarFieldsInCurrentRecord = 0;
           for (VarLengthColumn<?> columnReader : columns) {
    -        if ( !exitLengthDeterminingLoop ) {
    -          exitLengthDeterminingLoop = columnReader.determineSize(recordsReadInCurrentPass, lengthVarFieldsInCurrentRecord);
    +        if (!exitLengthDeterminingLoop) {
    +          exitLengthDeterminingLoop =
    +              columnReader.determineSize(recordsReadInCurrentPass, lengthVarFieldsInCurrentRecord);
             } else {
               break;
             }
           }
           // check that the next record will fit in the batch
    -      if (exitLengthDeterminingLoop || (recordsReadInCurrentPass + 1) * parentReader.getBitWidthAllFixedFields() + totalVariableLengthData
    -          + lengthVarFieldsInCurrentRecord > parentReader.getBatchSize()) {
    +      if (exitLengthDeterminingLoop ||
    +          (recordsReadInCurrentPass + 1) * parentReader.getBitWidthAllFixedFields()
    +              + totalVariableLengthData + lengthVarFieldsInCurrentRecord > parentReader.getBatchSize()) {
    +        break;
    +      }
    +      for (VarLengthColumn<?> columnReader : columns) {
    +        columnReader.updateReadyToReadPosition();
    +        columnReader.currDefLevel = -1;
    +      }
    +      recordsReadInCurrentPass++;
    +      totalVariableLengthData += lengthVarFieldsInCurrentRecord;
    +    } while (recordsReadInCurrentPass < recordsToReadInThisPass);
    +
    +    return recordsReadInCurrentPass;
    +  }
    +
    +
    +  public long determineSizesParallel(long recordsToReadInThisPass ) throws IOException {
    +    boolean doneReading = false;
    +    int lengthVarFieldsInCurrentRecord = 0;
    +    boolean exitLengthDeterminingLoop = false;
    +    long totalVariableLengthData = 0;
    +    long recordsReadInCurrentPass = 0;
    +
    +    do {
    +    doneReading = readPagesParallel();
    +
    +    if (!doneReading) {
    +      lengthVarFieldsInCurrentRecord = 0;
    +      for (VarLengthColumn<?> columnReader : columns) {
    +        doneReading = columnReader.processPageData((int) recordsReadInCurrentPass);
    +        if(doneReading) {
    +          break;
    +        }
    +        lengthVarFieldsInCurrentRecord += columnReader.dataTypeLengthInBits;
    +        doneReading = columnReader.checkVectorCapacityReached();
    +        if(doneReading) {
    +          break;
    +        }
    +      }
    +    }
    +
    +    exitLengthDeterminingLoop = doneReading;
    +
    +      // check that the next record will fit in the batch
    +      if (exitLengthDeterminingLoop ||
    +          (recordsReadInCurrentPass + 1) * parentReader.getBitWidthAllFixedFields()
    +              + totalVariableLengthData + lengthVarFieldsInCurrentRecord > parentReader.getBatchSize()) {
             break;
           }
    -      for (VarLengthColumn<?> columnReader : columns ) {
    +      for (VarLengthColumn<?> columnReader : columns) {
             columnReader.updateReadyToReadPosition();
             columnReader.currDefLevel = -1;
           }
           recordsReadInCurrentPass++;
           totalVariableLengthData += lengthVarFieldsInCurrentRecord;
         } while (recordsReadInCurrentPass < recordsToReadInThisPass);
     
    +    return recordsReadInCurrentPass;
    +  }
    +
    +  public boolean readPagesParallel() {
    +
    +    boolean isDone = false;
    +    ArrayList<Future<Boolean>> futures = Lists.newArrayList();
    +    for (VarLengthColumn<?> columnReader : columns) {
    +      Future<Boolean> f = columnReader.readPageAsync();
    +      futures.add(f);
    +    }
    +    for (Future<Boolean> f : futures) {
    +      try {
    +        isDone = isDone || f.get().booleanValue();
    +      } catch (Exception e) {
    +        f.cancel(true);
    +        handleAndRaise(null, e);
    +      }
    +    }
    +    return isDone;
    +  }
    +
    +
    +  private void readRecordsSerial(long recordsReadInCurrentPass) {
         for (VarLengthColumn<?> columnReader : columns) {
           columnReader.readRecords(columnReader.pageReader.valuesReadyToRead);
         }
         for (VarLengthColumn<?> columnReader : columns) {
    -      columnReader.valueVec.getMutator().setValueCount((int) recordsReadInCurrentPass);
    +      columnReader.valueVec.getMutator().setValueCount((int)recordsReadInCurrentPass);
         }
    -    return recordsReadInCurrentPass;
    +  }
    +
    +  private void readRecordsParallel(long recordsReadInCurrentPass){
    +    ArrayList<Future<Integer>> futures = Lists.newArrayList();
    +    for (VarLengthColumn<?> columnReader : columns) {
    +      Future<Integer> f = columnReader.readRecordsAsync(columnReader.pageReader.valuesReadyToRead);
    +      futures.add(f);
    +    }
    +    for (Future f : futures) {
    +      try {
    +        f.get();
    +      } catch (Exception e) {
    +        f.cancel(true);
    --- End diff --
    
    Done


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---