[CALCITE-2091] Improve DruidQuery cost function, to ensure that EXTRACT gets pushed as an interval if possible

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.16.0
• Component/s: druid-adapter
• Assignee: Nishant Bangarwa
• Reporter: Nishant Bangarwa
• Created: 15/Dec/17 16:48
• Updated: 27/Feb/24 22:23
• Resolved: 28/Jan/18 00:58

Description

DruidAdapterIT testFilterTimestamp does not test translating of extract fn to interval.

This test was added to verify that the extract function is transformed to a date range on timestamp column. But in this it does not seem to be pushing extract to interval. This task is to check the root cause and fix this. I guess this is possibly an issue with cost computation not accounting for interval being queried.

 
/** Tests that conditions applied to time units extracted via the EXTRACT
   * function become ranges on the timestamp column
   *
   * <p>Test case for
   * <a href="https://issues.apache.org/jira/browse/CALCITE-1334">[CALCITE-1334]
   * Convert predicates on EXTRACT function calls into date ranges</a>. */
  @Test public void testFilterTimestamp() {
    String sql = "select count(*) as c\n"
        + "from \"foodmart\"\n"
        + "where extract(year from \"timestamp\") = 1997\n"
        + "and extract(month from \"timestamp\") in (4, 6)\n";
    final String explain = "DruidQuery(table=[[foodmart, foodmart]], "
        + "intervals=[[1900-01-09T00:00:00.000Z/2992-01-10T00:00:00.000Z]], "
        + "filter=[AND(=(EXTRACT(FLAG(YEAR), $0), 1997), OR(=(EXTRACT(FLAG(MONTH), $0), 4), "
        + "=(EXTRACT(FLAG(MONTH), $0), 6)))], groups=[{}], aggs=[[COUNT()]])";
    sql(sql)
        .explainContains(explain)
        .returnsUnordered("C=13500");
  }