Given that OI32 removes the contraint of 64k record size, are there any guidelines/provisos regarding whether it is better to have one biggish record/array rather than breaking record/array into several smaller ones from the perspective of:

a. I/O performance

b. Get_Property(Array)

c. Set_Property(Array)

d. Constaints in forms (I am assuming the previous OI16 limits regarding no. of objects in forms has been removed)

Any suggestions would be welcome

Thanks and regards,

Godfrey,

OI32 is so very much faster that OI16. Period. As the conversion is easy to accomplish from OI16 to OI32 it is best to run the upgrade on your application and then perform benchmarks.

Some benchmark data is network constrained, some video constrained, some disk constrained, and some memory-constrained. This mitigates tests with your own application to check.

At some point the numbers become meaningless and the users have to perceive a difference. In the case of one Australian payroll, the payroll process was so fast that the test staff restored from backup several times and repeated the tests, until they realised that OI32 was crunching the entire payroll in two seconds - previously a job lasting many minutes.

Hopefully you will have your own miraculous anecdotes to share.

Steve

The Sprezzatura Group

World Leaders in all things RevSoft

I concur with Steves observation, but add that the performance gain is greatest on a 'true' Win32 O/S such as Win2000/XP. That is on the Win98SE client platform the performance improvement of OI32 over OI16 is quite noticeable; on the WinXP Pro platform the improvment is - shall I say - quite significant.

Honestly, having worked with OI32 for a while, I can't imagine why a developer would persist in developing in 16bit unless constrained by client requirements.

View this thread on the forum...