Up at last

Well it has finally happened, finished and live (yipee).

After some heavy pressure from myself and the other developers, who felt it was more productive to go live with bugs than spend the next month or two going through the rounds of debugging and testing. Obviously there are no major functionality bugs, more a matter of fine tuning and data issues.

We now no longer have data conversion issues to worry about. We can still alter all our converted data, and easily identify new data. Shortly after going live, with many favourable comments I might add, we uncovered some more minor bugs that no amount of user testing would have uncovered.

Altogether a most satisfactory result I think. All major bugs seem to have been fixed and now performance tuning reigns supreme.

The new spatial features seem to be pulling their weight and haved added the most oomph to the site, way to go!

What's On

There are times when I just want to scream!!!

Just trying to match up venues with listings on a weekly basis is enough of a challenge. The level of challenge increases when you consider that the external data supplier may or may not alter their internal identifiers for venues. We have already encountered situations where identifiers have been reused on different venues.

Still, soldiering on, into the home stretch now. Suprisingly the whole thing is starting to come together and provides fairly decent performance. The new site will look grand, and will be different from any mapping site I've seen 'out there'.

The basic design of the database has stood up well to all the abuses that have been heaped on it as data not intended to be stored in this structure was squeezed into contortions to fit the design and still provide acceptable performance.

With the help of SDO_TUNE and a distinct lack of documentation (anywhere), I've managed to increase performance of many features within SPATIAL. Working from a position of no baseline performance to aim for, I've had to set and tune parameters based on feeling and observed performance and the results have been impressive.

Until the next time.........

Exstensible Madness Language

Can't help but admire the guys (and gals) who designed XML. They have made a beast of a nightmare come true. When it works, it works really well, but when it doesn't..., well then the hair pulling starts.

Oracle have provided us with some lovely tools to generate and post XML direct from a SQL query, and with some tinkering with the lesser known features we actually got the thing to work. Now all we have to do is repeat the same thing many times over for searches scattered throughout the system (the consistent scatter-gun approach).

I am seriously considering modularising the entire back-end and giving it a more generic class like structure, though how this will affect performance is in the hands of the gods. My first attempts have been promising (so far), using the database design and key data to auto generate queries etc.... I'll keep you posted.

As for the spatial side of things - well, performance was suffering, as was my downtime, until I found some useful hints from others in the same situation. They pointed me towards some valuable performance enhancing methods for getting the most out of Spatial.