Month: February 2008

The Importance of Profiling

I’ve worked as a software developer and worked with teams of software developers for around 10 years now, Many of those whom I’ve worked with have earned my trust and respect in relation to development and testing techniques. Frustratingly however it’s still with irritating regularity that I hear throw-away comments bourne of uncertainty and ignorance.

A couple of times now I’ve specifically been told that “GD makes my code go slow”. Now for those of you not in the know GD (actually specifically Lincoln Stein’s GD.pm in perl) is a wrapper around Tom Boutell’s most marvellous libgd graphics library. The combination of these two has always performed excellently for me and never been the bottleneck in any of my applications. The applications in question are usually database-backed web applications with graphics components for plotting genomic features or charts of one sort or another.

As any database-application developer will tell you, the database, or network connection to the database is almost always the bottleneck in an application or service. Great efforts are made to ensure database services scale well and perform as efficiently as possible, but even after these improvements are made they usually simply delay the inevitable.

Hence my frustration when I hear that “GD is making my (database) application go slow”. How? Where? Why? Where’s the proof? It’s no use blaming something, a library in this case, that’s out of your control. It’s hard to believe a claim like that without some sort of measurement.

So.. before pointing the finger, profile the code and make an effort to understand what the profiler is doing. In database applications profile your queries – use EXPLAIN, add indices, record SQL transcripts and time the results. Then profile the code which is manipulating those results.

Once the results are in of course, concentrate in the first instance on the parts with the most impact (e.g. 0.1 second off each iteration of a 1000x loop rather than 1 second from /int main/ ) – the low hanging fruit. Good programmers should be relatively lazy and speeding up code with the least amount of effort should be commonsense.

Great pieces of code

A lot of what I do day-to-day is related to optimisation. Be it Perl code, SQL queries, Javascript or HTML there are usually at least a couple of cracking examples I find every week. On Friday I came across this:

SELECT cycle FROM goldcrest WHERE id_run = ?

This query is being used to find the number of the latest cycles (between 1 and 37 for each id_run) in a near-real-time tracking system and is used several times whenever a run report is viewed.

EXPLAIN SELECT cycle FROM goldcrest WHERE id_run = 231;
  
+----+-------------+-----------+------+---------------+---------+---------+-------+--------+-------------+
| id | select_type | table     | type | possible_keys | key     | key_len | ref   | rows   | Extra       |
+----+-------------+-----------+------+---------------+---------+---------+-------+--------+-------------+
|  1 | SIMPLE      | goldcrest | ref  | g_idrun       | g_idrun |       8 | const | 262792 | Using where |
+----+-------------+-----------+------+---------------+---------+---------+-------+--------+-------------+

In itself this would be fine but the goldcrest table in this instance contains several thousand rows for each id_run. So, for id_run, let’s say, 231 this query happens to return approximately 588,000 rows to determine that the latest cycle for run 231 is the number 34.

To clean this up we first try something like this:

SELECT MIN(cycle),MAX(cycle) FROM goldcrest WHERE id_run = ?

which still scans the 588000 rows (keyed on id_run incidentally) but doesn’t actually return them to the user, only one row containing both values we’re interested in. Fair enough, the CPU and disk access penalties are similar but the data transfer penalty is significantly improved.

Next I try adding an index against the id_run and cycle columns:

ALTER TABLE goldcrest ADD INDEX(id_run,cycle);
Query OK, 37589514 rows affected (23 min 6.17 sec)
Records: 37589514  Duplicates: 0  Warnings: 0

Now this of course takes a long time and, because the tuples are fairly redundant, creates a relatively inefficient index, also penalising future INSERTs. However, casually ignoring those facts, our query performance is now radically different:

EXPLAIN SELECT MIN(cycle),MAX(cycle) FROM goldcrest WHERE id_run = 231;
  
+----+-------------+-------+------+---------------+------+---------+------+------+------------------------------+
| id | select_type | table | type | possible_keys | key  | key_len | ref  | rows | Extra                        |
+----+-------------+-------+------+---------------+------+---------+------+------+------------------------------+
|  1 | SIMPLE      | NULL  | NULL | NULL          | NULL |    NULL | NULL | NULL | Select tables optimized away |
+----+-------------+-------+------+---------------+------+---------+------+------+------------------------------+
  
SELECT MIN(cycle),MAX(cycle) FROM goldcrest WHERE id_run = 231;
+------------+------------+
| MIN(cycle) | MAX(cycle) |
+------------+------------+
|          1 |         37 |
+------------+------------+
  
1 row in set (0.01 sec)

That looks a lot better to me now!

Generally I try to steer clear of the mysterious internal workings of database engines, but with much greater frequency come across examples like this:

sub clone_type {
  my ($self, $clone_type, $clone) = @_;
  my %clone_type;

  if($clone and $clone_type) {
    $clone_type{$clone} = $clone_type;
    return $clone_type{$clone};
  }

  return;
}

Thankfully this one’s pretty quick to figure out – they’re usually *much* more convoluted, but still.. Huh??

Pass in a clone_type scalar, create a local hash with the same name (Argh!), store the clone_type scalar in the hash keyed at position $clone, then return the same value we just stored.

I don’t get it… maybe a global hash or something else would make sense, but this works out the same:

sub clone_type {
  my ($self, $clone_type, $clone) = @_;

  if($clone and $clone_type) {
    return $clone_type;
  }
  return;
}

and I’m still not sure why you’d want to do that if you have the values on the way in already.

Programmers really need to think around the problem, not just through it. Thinking through may result in functionality but thinking around results in both function and performance which means a whole lot more in my book, and incidentally, why it seems so hard to hire good programmers.