Hi,
du to the problem with TOracleDataset and "Out of memory" when the query fetch a lot of VARCHAR2(4000) fields. As I understand TOracleDataset allocated 4000 characters for every field even if there is only a few characters of actual data(?)

I have seen the suggestion to set the dataset Unidirectional, but that is not an option since we use Master/Detail and we then keep getting a (show)message like "Record 33 not accessible in Unidirectional mode".

Do you have any suggestion to get around the "out of memory" issue?

You can perhaps set TOracleDataSet.QueryAllRecords to False, so that the dataset will only fetch the records when they are needed.

We do set QueryAllRecord to false. But as soon as the user start to move through the dataset the memory consumption builds up. And if the user choose to do an export to EXCEL we need to scroll through the dataset and memory consuption hits the roof.
Off course we only need 25-50 records in memory at any given time - so how do we free up the memory for the records we don't need? Is it possible?

Does a user really scroll through 10,000 or 100,000 records? It doesn't seem very useful. Usually a simple QBE implementation will help a user to quickly find the records he or she needs without any memory overhead.

To export to Excel you can use the UniDirectional option.

Hi, sorry for late answer - but I don't get notifications when you reply...

As I worte we have a implementation using master/detail - we do need this. No way around.

A customer can have 100/1000/10000 (seldom 100000 - but still possible) (contracts/invoices/etc). The user rarley scroll through 10.000 or 100.000 but as I wrote in my first post we can not scroll at all when using undirectional dataset. Or is there somehting we have misunderstood?

How do we user master/detail with undirection dataset?