asterixdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Yingyi Bu <buyin...@gmail.com>
Subject Re: Does Projection affect count() performance?
Date Sat, 01 Oct 2016 16:43:07 GMT
Wail,

      Can you attach the query plan for query 1?
      I tried
       count( for $x in dataset beers
         return $x
       )

      and got the following plan, which seems OK:
-- DISTRIBUTE_RESULT  |UNPARTITIONED|
  exchange
  -- ONE_TO_ONE_EXCHANGE  |UNPARTITIONED|
    aggregate [$$5] <- [function-call: asterix:agg-sum, Args:[%0->$$8]]
    -- AGGREGATE  |UNPARTITIONED|
      exchange
      -- RANDOM_MERGE_EXCHANGE  |PARTITIONED|
        aggregate [$$8] <- [function-call: asterix:agg-count,
Args:[%0->$$0]]
        -- AGGREGATE  |PARTITIONED|
          project ([$$0])
          -- STREAM_PROJECT  |PARTITIONED|
            exchange
            -- ONE_TO_ONE_EXCHANGE  |PARTITIONED|
              data-scan []<-[$$6, $$0, $$7] <- Default:beers
              -- DATASOURCE_SCAN  |PARTITIONED|
                exchange
                -- ONE_TO_ONE_EXCHANGE  |PARTITIONED|
                  empty-tuple-source
                  -- EMPTY_TUPLE_SOURCE  |PARTITIONED|


Best,
Yingyi

On Sat, Oct 1, 2016 at 9:25 AM, Mike Carey <dtabass@gmail.com> wrote:

> Sounds like there is a new materialization bug there.....  Please file a
> JIRA issue (and we'll need a query plan test case to keep it from breaking
> again).
>
>
>
> On 10/1/16 2:01 AM, Wail Alkowaileet wrote:
>
>> Hi,
>>
>> I know that early projections will enhance the performance.
>> I just noticed something:
>>
>> 1- returning the whole tuple
>> count( for $x in dataset Tweets
>> return $x
>> )
>>
>> => Throws an exception Java heap exceeded. (The heap-size is less than the
>> sum of AsterixDB configured memory ... so it's not a problem).
>>
>> 2- However, returning one field
>> count( for $x in dataset Tweets
>> return $x.id
>> )
>>
>> => Worked just fine.
>>
>> I'm just wondering, does the projection in count() affects its
>> performance ?
>>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message