database/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-08-01 03:47:19 +03:00
Code Activity

Merge spetrunia@bk-internal.mysql.com:/home/bk/mysql-5.1-opt

Browse Source 
into  mysql.com:/home/psergey/mysql-5.1-bug22393


mysql-test/t/range.test:
  Auto merged
sql/sql_select.cc:
  Auto merged
mysql-test/r/range.result:
  Manual merge
This commit is contained in:
unknown
2006-09-19 21:14:37 +04:00
parent 86b4ca8480 5e34af3240
commit d752b1a8da
3 changed files with 64 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
sql/sql_select.cc
View File

@ -3717,7 +3717,30 @@ best_access_path(JOIN *join,
{
/* Check if we have statistic about the distribution */
if ((records= keyinfo->rec_per_key[max_key_part-1]))
{
/*
Fix for the case where the index statistics is too
optimistic: If
(1) We're considering ref(const) and there is quick select
on the same index,
(2) and that quick select uses more keyparts (i.e. it will
scan equal/smaller interval then this ref(const))
(3) and E(#rows) for quick select is higher then our
estimate,
Then
We'll use E(#rows) from quick select.
Q: Why do we choose to use 'ref'? Won't quick select be
cheaper in some cases ?
TODO: figure this out and adjust the plan choice if needed.
*/
if (!found_ref && table->quick_keys.is_set(key) && // (1)
table->quick_key_parts[key] > max_key_part && // (2)
records < (double)table->quick_rows[key]) // (3)
records= (double)table->quick_rows[key];
tmp= records;
}
else
{
/*