You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues-all@impala.apache.org by "Tim Armstrong (JIRA)" <ji...@apache.org> on 2018/10/30 17:42:00 UTC
[jira] [Updated] (IMPALA-2743) Propagate IS NOT NULL to scans from
cols in inner joins
[ https://issues.apache.org/jira/browse/IMPALA-2743?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Tim Armstrong updated IMPALA-2743:
----------------------------------
Issue Type: Improvement (was: Bug)
> Propagate IS NOT NULL to scans from cols in inner joins
> -------------------------------------------------------
>
> Key: IMPALA-2743
> URL: https://issues.apache.org/jira/browse/IMPALA-2743
> Project: IMPALA
> Issue Type: Improvement
> Components: Frontend
> Affects Versions: Impala 2.2, Impala 2.3.0
> Reporter: Mostafa Mokhtar
> Priority: Minor
> Labels: performance, planner, ramp-up
>
> Columns involved inner joins and filters should have a "IS NOT NULL" filter pushed to the ScanNode. Other operators reading the value for the column shouldn't be re-checking if the value is null again.
> When using Parquet file format ideally we should be using the stats to short circuit these filters.
> Currently whenever a slot is read by any of the operators SlotRef checks if the value is NULL which adds unnecessary branches and instructions.
> {code}
> BigIntVal SlotRef::GetBigIntVal(ExprContext* context, TupleRow* row) {
> DCHECK_EQ(type_.type, TYPE_BIGINT);
> Tuple* t = row->GetTuple(tuple_idx_);
> if (t == NULL || t->IsNull(null_indicator_offset_)) return BigIntVal::null();
> return BigIntVal(*reinterpret_cast<int64_t*>(t->GetSlot(slot_offset_)));
> }
> {code}
> {code}
> impalad!impala::Tuple::IsNull
> impalad!impala::SlotRef::GetBigIntVal - [Unknown]
> impalad!impala::SlotRef::GetBigIntVal - [Unknown]
> impalad!impala::ExprContext::GetValue+0x13b - [Unknown]:[Unknown]
> impalad!impala::HashTableCtx::EvalRow+0xb7 - [Unknown]:[Unknown]
> impalad!impala::PartitionedHashJoinNode::Partition::BuildHashTableInternal<(bool)0>+0x372 - [Unknown]:[Unknown]
> impalad!impala::PartitionedHashJoinNode::Partition::BuildHashTable+0xd - [Unknown]:[Unknown]
> impalad!impala::PartitionedHashJoinNode::BuildHashTables+0xa0 - [Unknown]:[Unknown]
> impalad!impala::PartitionedHashJoinNode::ProcessBuildInput+0xc71 - [Unknown]:[Unknown]
> impalad!impala::PartitionedHashJoinNode::ConstructBuildSide+0x106 - [Unknown]:[Unknown]
> impalad!impala::BlockingJoinNode::BuildSideThread+0x7f - [Unknown]:[Unknown]
> impalad!impala::Thread::SuperviseThread+0x1b9 - [Unknown]:[Unknown]
> impalad!boost::detail::thread_data<boost::_bi::bind_t<void, void (*)(std::string const&, std::string const&, boost::function<void (void)>, impala::Promise<long>*), boost::_bi::list4<boost::_bi::value<std::string>, boost::_bi::value<std::string>, boost::_bi::value<boost::function<void (void)>>, boost::_bi::value<impala::Promise<long>*>>>>::run+0x7f - [Unknown]:[Unknown]
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-all-unsubscribe@impala.apache.org
For additional commands, e-mail: issues-all-help@impala.apache.org