Bug#48846: Too much time spent in ha_partition::records_in_range if not able to prune

Problem was that ha_partition::records_in_range called
records_in_range for all non pruned partitions, even if
an estimate should be given.

Solution is to only use 1/3 of the partitions (up to 10) for
records_in_range and estimate the total from this subset.
(And continue until a non zero return value from the called
partitions records_in_range is given, since 0 means no rows
will match.)

sql/ha_partition.cc:
  Bug#48846: Too much time spent in ha_partition::records_in_range if not able to prune
  
  estimate_rows_upper_bound and records_in_range are very similar
  (the only difference is the function and its parameters to use)
  so I created a common function for this.
  
  Since these calls from the optimizer are only estimates, it is
  not neccesary to call them for every partition, it can use
  a much smaller subset of the used partitions instead,
  which improves performance for selects.
sql/ha_partition.h:
  Bug#48846: Too much time spent in ha_partition::records_in_range if not able to prune
  
  Added two private functions to help some
  optimizer calls.

1 files changed, 12 insertions, 0 deletions

diff --git a/sql/ha_partition.h b/sql/ha_partition.h
index c08b1f77eca..9f6d9e0a5ba 100644
--- a/sql/ha_partition.h
+++ b/sql/ha_partition.h
@@ -547,6 +547,18 @@ public:
      -------------------------------------------------------------------------
   */
 
+private:
+  /*
+    Helper function to get the minimum number of partitions to use for
+    the optimizer hints/cost calls.
+  */
+  void partitions_optimizer_call_preparations(uint *num_used_parts,
+                                              uint *check_min_num,
+                                              uint *first);
+  ha_rows estimate_rows(bool is_records_in_range, uint inx,
+                        key_range *min_key, key_range *max_key);
+public:
+
   /*
     keys_to_use_for_scanning can probably be implemented as the
     intersection of all underlying handlers if mixed handlers are used.