36385[TASK] Add index key to sys_category_record_mm
Commit message
Review
-
Code review CR +2
- Anja Leichsenring +2
- Mathias Schreiber +1
- Jan Helke +1
- Dirk Jüttner +1
-
Verified V +2
- Anja Leichsenring +2
- Mathias Schreiber +1
- Jan Helke +1
- Dirk Jüttner +1
Issues 1
Files 1
Review log 27
Show 2 earlier messages
-
Jan Helke voted on patch set 1 CR +1 V +1 01/30/15, 13:38
-
Jan Helke updated the commit message in patch set 2 01/30/15, 13:40
-
Jan Helke voted on patch set 2 CR +1 V +1 01/30/15, 13:41
-
Göran Bodenschatz voted on patch set 2 CR +1 01/30/15, 13:41
-
Christian Kuhn commented on patch set 2 CR -1 01/30/15, 13:59
-1 for now asking for the actual queries fired to the table to see if we could maybe remove one of the other indexes.ext_tables.sql:708i wonder if this key could be removed then. Which queries are actually fired to these table? -
Bodo Eichstädt commented on patch set 1 01/30/15, 14:52
Can't proof the absence of uid_foreign + tablenames constraints query, but actually mm funcs provide actually ANY combination of uid_foreign AND/OR fieldname AND/OR tablenames. Contrary the presence of ALL three constraints id easy to find. -
Stefan Froemken commented on patch set 2 01/30/15, 22:41
With help of sfmysqlreport I have a set of over 10.000 queries and all queries with sys_category_record_mm have uid_foreign, filename and tablename defined. I would prefer to add sorting_foreign as forth element to the key, because of ordering -
Stefan Froemken uploaded patch set 3 01/30/15, 22:48
-
Stefan Froemken uploaded patch set 4 01/30/15, 23:46
-
Stefan Froemken commented on patch set 4 01/30/15, 23:54
What do you think about that solution? In previous version we have varchar(255) with utf-8 collation. Sorry, but I have never seen chinese characters as column names. Or do you have seen russia characters in tablenames? I don't think so. So why still using varchar. Don't forget: In case of utf8 EACH letter needs 3 bytes in database. With tablename AND fieldname in key we need 1530 bytes. By default MySQL allows only 1000 bytes each key. So switching to tinyblob allows us to still save 255 ASCII letters and have both fields in key. Further EXPLAIN tolds us that filesort will not be needed anymore which increases performance. -
Oliver Klee commented on patch set 4 01/31/15, 11:37
ext_tables.sql:699I think this should still be some text type the type matches the content. -
Stefan Froemken commented on patch set 4 01/31/15, 12:24
ext_tables.sql:699ReplySounds good. But in that case I would prefer to allow COLLATION and CHARACTER SET in these sql-files. Then we can use a MySQL valid ASCII charcter set. -
Anja Leichsenring updated the commit message in patch set 5 01/31/15, 12:39
-
Bodo Eichstädt commented on patch set 5 CR +1 01/31/15, 14:12
I agree with Oliver, char is more convenient for the quick peek with phpmyadmin -
Oliver Klee commented on patch set 5 CR -1 01/31/15, 14:21
Bodo, will you push a new version? -
Göran Bodenschatz voted on patch set 5 CR +1 01/31/15, 14:22
-
Zbigniew Jacko commented on patch set 5 02/01/15, 09:46
When I compare database execute: "Database update failed Error: BLOB/TEXT column 'tablenames' used in key specification without a key length" -
Zbigniew Jacko commented on patch set 5 02/01/15, 09:53
Ok, but when I did 'step by step' changes it works well. -
Mathias Schreiber commented on patch set 5 CR -2 V +1 02/01/15, 09:56
Arguably +1 on the index, -2 on the tinyblob unless properly explained -
Dirk Jüttner uploaded patch set 6 04/30/15, 20:20
-
Mathias Schreiber voted on patch set 6 CR +1 V +1 04/30/15, 20:27
-
Jan Helke voted on patch set 6 CR +1 V +1 04/30/15, 20:33
-
Dirk Jüttner voted on patch set 6 CR +1 V +1 04/30/15, 20:34
-
Anja Leichsenring voted on patch set 6 CR +2 V +2 04/30/15, 20:35
-
Anja Leichsenring merged patch set 6 04/30/15, 20:35