- architecture somewhat old
- primarily optimized towards oltp
- not many low hanging fruits
- hashtables
  - hash based aggregation
  - bitmpscans
- expression evaluation
  - "mini" Virtual-Machine
  - possibly JITable
- incrementally changing executor to batch tuples
  - show plan
  - explain flow
  - explain batches


Aggregate  (actual time=...6691.528 rows=1)
  ->  Bitmap Heap Scan on lineitem  (actual time=..5195.704 rows=9113219)
        Recheck Cond: ((l_shipdate >= '1995-01-01'::date) AND (l_shipdate <= '1996-12-31'::date))
        Heap Blocks: exact=585542
        ->  Bitmap Index Scan on i_l_shipdate  (actual time=..4429.997 rows=9113219)
              Index Cond: ((l_shipdate >= '1995-01-01'::date) AND (l_shipdate <= '1996-12-31'::date))
Execution time: 5196.203 ms

Aggregate  (actual time=..3341.864 rows=1)
  ->  Bitmap Heap Scan on lineitem  (actual time=..2569.560 rows=9113219)
        Recheck Cond: ((l_shipdate >= '1995-01-01'::date) AND (l_shipdate <= '1996-12-31'::date))
        Heap Blocks: exact=585542
        ->  Bitmap Index Scan on i_l_shipdate  (actual time=..951.016 rows=9113219)
              Index Cond: ((l_shipdate >= '1995-01-01'::date) AND (l_shipdate <= '1996-12-31'::date))
Execution time: 3343.101 ms



EXPLAIN (ANALYZE, TIMING OFF) SELECT l_partkey, SUM(l_extendedprice) FROM lineitem GROUP BY 1 ORDER BY 2 DESC LIMIT 10;

Limit  (actual rows=10 loops=1)
  ->  Sort  (actual rows=10 loops=1)
        Sort Key: (sum(l_extendedprice)) DESC
        Sort Method: top-N heapsort  Memory: 25kB
        ->  HashAggregate  (cost=1038925.33..1049377.88 rows=1045255 width=12) (actual rows=1000000 loops=1)
              Group Key: l_partkey
              ->  Seq Scan on lineitem  (cost=0.00..888926.22 rows=29999822 width=12) (actual rows=29999795 loops=1)
Execution time: 17803.043 ms

Limit  (cost=1071965.46..1071965.49 rows=10 width=12) (actual rows=10 loops=1)
  ->  Sort  (cost=1071965.46..1074578.60 rows=1045255 width=12) (actual rows=10 loops=1)
        Sort Key: (sum(l_extendedprice)) DESC
        Sort Method: top-N heapsort  Memory: 25kB
        ->  HashAggregate  (cost=1038925.33..1049377.88 rows=1045255 width=12) (actual rows=1000000 loops=1)
              Group Key: l_partkey
              ->  Seq Scan on lineitem  (cost=0.00..888926.22 rows=29999822 width=12) (actual rows=29999795 loops=1)
Execution time: 12444.961 ms



SELECT l_partkey, SUM(l_extendedprice) FROM lineitem GROUP BY 1 ORDER BY 2 DESC LIMIT 10;

Limit  (actual time=..18529.047 rows=10)
  ->  Sort  (actual time=..18529.045)
        Sort Key: (sum(l_extendedprice)) DESC
        Sort Method: top-N heapsort  Memory: 25kB
        ->  HashAggregate  (actual time=..18427.389 rows=1000000)
              Group Key: l_partkey
              ->  Seq Scan on lineitem  (actual time=..3106.676 rows=29999795)
Execution time: 18536.564 ms

Limit  (actual time=..14087.096 rows=10)
  ->  Sort  (actual time=..14087.093 rows=10)
        Sort Key: (sum(l_extendedprice)) DESC
        Sort Method: top-N heapsort  Memory: 25kB
        ->  HashAggregate  (actual time=..14047.655 rows=1000000)
              Group Key: l_partkey
              ->  Seq Scan on lineitem  (actual time=..3136.398 rows=29999795)
Execution time: 14102.279 ms



SELECT count(*) FROM lineitem WHERE l_extendedprice > 1000;

ExecMakeFunctionResultNoSets()
  check_stack_depth()
  ExecEvalFuncArgs()
    ExecEvalScalarVarFast()
    ExecEvalConst()
  float8gt()