It looks like we were causing this behaviour, instead of it being how Class::DBI operates.

By default DATE fields in Oracle are formatted for year, month, and day even though the field includes hour, minute and second. In order to get the complete date and time our base class changes the default formatting with a line similar to:

__PACKAGE__->db_Main->do(
  "ALTER SESSION SET NLS_DATE_FORMAT = 'DD:MM:YYYY HH24:MI:SS'"
);

It is this line that triggers a connection on module load. If it is removed then the classes can be loaded even if the database is dead. Connection is now only attempted on first action which is a much more approriate time.

I’m not sure what the documentation said about it when we put the line in, but Setting Session Globals and Working With Oracle Date Fields on the Class::DBI wiki warn about using it in a mod_perl environment and that it will initiate a connection when the modules is used.

Over the next few days the commercial product that this database is used with is being upgraded. During this time the database will be unavailable for use so we revisited the issue of how to get the classes to load cleanly. These investigations found the root cause and we now have a solution; alter the base class so that the date formatting is altered whenever a connection is initialized (using information from how to use multiple databases), not when the class is loaded.

Now, although we have solved this issue, having to solve it wasn’t trivial as we were required to dig through some of the Class::DBI and Ima::DBI internals to mimic its behaviour.

• As I had seen Scott talk about Zaltana before (at OSDC in the past, at OSDClub, and at perl mongers) I decided that I would take my time getting in and skip his keynote. I ended up taking more time than I planned and missed the next talk slot.
• Adam took up two half hour slots with his presentation of The Portable Image Testing Architecture: Rediculously Large Scale Testing. In a nutshell this is a system of virtual machines running a variety of operating systems with a variety of perl versions. Perl modules are then fed to these machines in order to obtain test results. This is what CPAN testers will be replaced with.
• Another Scott (actually one that I see every day at work) presented on Usability, user-centered design (UCD) and FOSS. Although I had been exposed to the material before it was still an excellent presentation which was represented by Scott receiving a prize for the best talk later in the day.
• As usual the lightning talk session was the highlight of the day with many excellent talks and some good talks. Cog has even made available the slides from his lightning talk which summarise the entire conference

Despite excellent talks and more people than last year I could not shake the impression that the conference was not as good as the previous years. This was partly confirmed by the announcement that next year it will NOT be held in Melbourne which would expose a new audience. The question is whether I would travel to attend…

See also:

• Journal of cog – OSDC – day 3
• Flickr: Photos tagged with osdcmelb2006
• PJF’s Pages – Journal – What’s new in Perl 5.10; OSDC, day 3 + aftermath

Tomorrow on the other hand is looking great.

See also:

• PJF’s Pages – Journal – OSDC day 2
• Journal of cog – OSDC – day 2

For now I am relaxing and will post a recap later tonight.

Anyway … today was the first day of OSDC 2006 with the following highlights:

• The opening keynote, Free software – A look back, a look ahead, by Randall Schwartz was a great start and included a run down of the various open source licences and examples of how Randall and others have been able to make a living from them. Despite them being ‘free’.
• cog’s first talk of the day, Perl White Magic, ran through some of the many command line switches and special variables that can be used to reduce the size of Perl scripts, in particular when writing one liners on the command line.
• In contrast to his first talk, cog’s second talk on obfuscation and golfing was the opposite of his earlier talk – how command line switches, special variables and other syntactic magic of Perl can be used for evil.
• Paul Fenwick gave us a rundown of the new features available in Perl 5.10 5.9.4. It is interesting to see Perl 6 features being incorporated. Jokes were made that by the time that Perl 6 comes out the features will already be available and in use.
• Adam Kennedy decided to buck the trend and give a talk on how things do not always go to plan. After reflecting that conference talks are (almost) always about new or succesful projects there is not exposure to the failures he went on to decribe a number of common pitfalls that are experienced.
• The final talk from cog covered ninety modules from the Acme namespace in twenty minutes (it should have taken thirty so he will look for some more to describe). What was scary about this was that some of them (for example one to allow C style comments) may now be useful to members of the audience. It was interesting to note that many of the mentioned modules were from Australian authors such as Damian , Paul and Adam.
• Above all of this was Damian’s rendition of The Da Vinci Codebase. This is a magnificent parody of The Da Vinci Code with the main protagonist, Dr Damian Conway and expert on symbology in computer languagess, supported by a variety of characters in Paris, Texas unravelling the clues that lead to the villan, the DMCA.

Although not part of the formal program, and facilitated by alcohol, were a variety of interesting discussions. On in particular led to the proposal of Acme::Playmate::Object. An extension of the Acme::Playmate module which allows you to obtain the vital statistics of Playboy playmates with Data::Vitals which is a class build for a modelling website that represents the vital statistics of fashion models and allows easy conversion between metric and imperial. The expectation is that Adam will have Acme::Playmate::Object (the objectification of playmates) ready to present as a lightning talk tomorrow.

See also:

• PJF’s Pages – Journal – OSDC-2006, Day 1
• Journal of cog – OSDC: day 1

I am also undecided on how I am going to get there. Last year I caught the train from Glen Waverley to Richmond and then another back to Caulfield which took around fifty minutes. I have been playing with the Metlink Journey Planner which tells me that it would be five minutes quicker to get off the train at Tooronga to catch a bus to Caulfield if I catch a certain train.

In a sense this highlights the issue of public transport in Melbourne being focused on travelling to the city. East Malvern station on the Glen Waverley line is actually pretty close to Caulfield station on the Packenham/Cranbourne/Frankston lines and has plenty of options including trains, buses and trams that can get you to the city. However the options to travel across to Caulfield are limited with no direct routes, they all go a fair distance out of the way…

Actually that statement is a bit misleading. Every available minute for the past two weeks except for a half dozen hours has been spent procrastinating instead of working on my paper. It is amazing what I would find myself doing.

Now I have three weeks until the conference. Three weeks of procrastination regarding working on my presentation…

I have made it available to anyone who cares along with my paper from two years ago.