BUG#27927:Partition pruning not optimal with TO_DAYS and YEAR functions

- Introduced val_int_endpoint() function which converts between func 
  argument intervals and func value intervals for monotonic functions.
- Made partition interval analyzer use part_expr->val_int_endpoint()
  to check if the edge values should be included.


mysql-test/r/partition_pruning.result:
  BUG#27927: Partition pruning not optimal with TO_DAYS and YEAR functions
  - Testcase
mysql-test/t/partition_pruning.test:
  BUG#27927: Partition pruning not optimal with TO_DAYS and YEAR functions
  - Testcase
sql/item.cc:
  BUG#27927: Partition pruning not optimal with TO_DAYS and YEAR functions
  - Added Item_field::val_int_endpoint() implementation
sql/item.h:
  BUG#27927: Partition pruning not optimal with TO_DAYS and YEAR functions
  - Added Item::val_int_endpoint() which converts intervals from argument
    space to function value space for unary monotonic functions.
sql/item_timefunc.cc:
  BUG#27927: Partition pruning not optimal with TO_DAYS and YEAR functions
  - Added val_int_endpoint() for TO_DAYS and YEAR functions.
sql/item_timefunc.h:
  BUG#27927: Partition pruning not optimal with TO_DAYS and YEAR functions
  - Added val_int_endpoint() for TO_DAYS and YEAR functions.
sql/partition_info.h:
  BUG#27927: Partition pruning not optimal with TO_DAYS and YEAR functions
  - Removed partition_info::range_analysis_include_bounds as it is no longer 
    needed.
sql/sql_partition.cc:
  BUG#27927: Partition pruning not optimal with TO_DAYS and YEAR functions
  - Make partition interval analyzer use part_expr->val_int_endpoint() to 
    check if the edge values should be included.
This commit is contained in:
unknown 2007-09-14 14:18:42 +04:00
commit 4aaabb06c0
8 changed files with 182 additions and 27 deletions

View file

@ -139,20 +139,6 @@ public:
*/
get_partitions_in_range_iter get_subpart_iter_for_interval;
/*
Valid iff
get_part_iter_for_interval=get_part_iter_for_interval_via_walking:
controls how we'll process "field < C" and "field > C" intervals.
If the partitioning function F is strictly increasing, then for any x, y
"x < y" => "F(x) < F(y)" (*), i.e. when we get interval "field < C"
we can perform partition pruning on the equivalent "F(field) < F(C)".
If the partitioning function not strictly increasing (it is simply
increasing), then instead of (*) we get "x < y" => "F(x) <= F(y)"
i.e. for interval "field < C" we can perform partition pruning for
"F(field) <= F(C)".
*/
bool range_analysis_include_bounds;
/********************************************
* INTERVAL ANALYSIS ENDS
********************************************/