Add support for dynamic filtering

fact_table - large fact table, partitioned on date_key d_date - small dimension table, with significant correlation on it’s columns and date_key.

Currently for the following query:

SELECT 
    count(*)
FROM 
    fact_table a 
JOIN d_date b 
ON a.date_key = b.date_key 
WHERE 
    b.year = 2017;

larger probe table is fully scanned. When presto after scaning build table (right) could extract information that for b.year = 2017 there are only small number of matching b.date_key values. This information could be used to narrow down the table scan. In other words, above query could be dynamically “rewritten” to equivalent of:

SELECT 
    count(*)
FROM 
    fact_table a 
JOIN d_date b 
ON a.date_key = b.date_key 
WHERE a.date_key IN (20170101,20170102, etc...)

[himanshukhantwal]

@maciejgrzybek @arhimondr Do we know if anyone working on this feature? Thanks!

[arhimondr]

@tastyqbit Presto Foundation actively works on getting this feature in (https://github.com/prestosql/presto/pull/91). Once it is there - we will cherry-pick it.