C[omp]ute

Welcome to my blog, which was once a mailing list of the same name and is still generated by mail. Please reply via the "comment" links.

Always interested in offers/projects/new ideas. Eclectic experience in fields like: numerical computing; Python web; Java enterprise; functional languages; GPGPU; SQL databases; etc. Based in Santiago, Chile; telecommute worldwide. CV; email.

Personal Projects

Choochoo Training Diary

Last 100 entries

MySql Efficiency, Security

From: "andrew cooke" <andrew@...>

Date: Mon, 13 Mar 2006 22:32:33 -0400 (CLT)

Rewrote the database for my current project to ensure all queries are
using indices.  It was a fair amount of work - partly because I had
experimented with some ideas that didn't work out (data in statements
rather than tables - even though it was very fixed data, it blocked
optimisation) and some ideas that MySql cannot optimise through (I'm not
100% clear it's MySql's fault, but I think that there are cases where the
view would propogate index information better if it were a merge, but it
can't be because of the definition, which leads to a temporary table).

Anyway, end result is much more clarity and efficiency.

And with the database cleaned up, security can work, which means more
learning - it's not as easy as I thought to find a simple set of
primitives that give you the control you want without a lot of complexity.

Still, progress.  I really think this could be done within the year.

Andrew

Useful SQL Commands

From: "andrew cooke" <andrew@...>

Date: Tue, 14 Mar 2006 07:14:39 -0400 (CLT)

The basic command for tuning selects is "describe select", so to check
that indices are being used you execute, for example:

  describe select * from table where a = b;

Note the a = b condition - this needs to be what will be actuallt used (if
you have no condition, for example, you're going to get an ALL since the
whole table is selected).

This prints a table (make your window wide) that includes the column
"type" - the contents of that column should NOT include "ALL".

I can't work out how to see what the view algorithm is - it seems that
"show create view name" should do this, ut it always gives "algorithm =
default".

An Example Using Group By

From: "andrew cooke" <andrew@...>

Date: Tue, 14 Mar 2006 08:28:23 -0400 (CLT)

Here's an example that requires a full table scan even though, as far sa i
can see, the underlying indices are sufficient:


drop table simple;
drop table duplicates;
drop view grouped;

create table simple (
  idx integer not null auto_increment primary key,
  val integer not null unique
);

insert into simple (val)
values (1), (2), (3), (4), (5), (6);

create table duplicates (
  val integer not null,
  dup integer not null,
  primary key (val, dup)
);

insert into duplicates (val, dup)
values (1,1), (2,2), (3,3), (3,30), (4,4), (4,40);

create view grouped as
  select s.idx, s.val, max(d.dup)
    from simple as s, duplicates as d
    where s.val = d.val
    group by idx;

-- this is slow, because grouping forces use of a temp table

select * from grouped where idx = 1;
describe select * from grouped where idx = 1;

-- this is equivalent, but faster

select s.idx, s.val, max(d.dup)
    from simple as s, duplicates as d
    where s.val = d.val
      and idx = 1
    group by idx;
describe select s.idx, s.val, max(d.dup)
    from simple as s, duplicates as d
    where s.val = d.val
      and idx = 1
    group by idx;

You can see the speed difference from the "describe" output.

Using the view:

+----+-------------+------------+-------+
| id | select_type | table      | type  |
+----+-------------+------------+-------+
|  1 | PRIMARY     | <derived2> | ALL   | <--- BAD!
|  2 | DERIVED     | s          | index |
|  2 | DERIVED     | d          | ref   |
+----+-------------+------------+-------+

Using the direct select:

+----+-------------+-------+-------+
| id | select_type | table | type  |
+----+-------------+-------+-------+
|  1 | SIMPLE      | s     | const |
|  1 | SIMPLE      | d     | ref   |
+----+-------------+-------+-------+

Comment on this post