Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Not a Bug
-
5.5.33a
-
None
-
None
-
None
Description
it looks like code for aria/hash index is rewritten, while the same for memory|myisam /hash is the same not so good.
tests are made for tables of:
engine = memory, table names are ttmp_mem_*
engine = aria, while the tables are actually on memdisk: ttmp_aria_*
engine = myisam, while tables are actually on memdis: ttmp_myisam_*
engine = aria, tables on SSD disk: tmp_aria_*
engine = myisam, tables on SSD disk: tmp_myisam_*
so update of randomly filled table of ~0.5mln rows, which affect those index , takes different times:
ttmp_mem_btree: 0.30 sec
ttmp_mem_hash: 38.28 sec # THAT IS TIME I SUPPOSE TO BE A BUG
ttmp_myisam_hash: 38.30 sec
ttmp_myisam_btree: 1.30 sec
ttmp_aria_hash: 1.22 sec # BECAUSE OF THIS
ttmp_aria_btree: 1.63 sec
tmp_myisam_hash: 38.09 sec
tmp_myisam_btree: 1.14 sec
tmp_aria_hash: 1.37 sec
tmp_aria_btree: 1.40 sec
sql-script to create+update tables in attach