Custom base class, key generator, custom mapping, seeing sql on error, email support

A few questions:

1. Is it possible to use custom base class?

2. I would like to create a custom identity generator similar to keytable, which would work per table field instead of per database. I guess custom base class overriding GeneratedId would be the correct way? How can I get the table and field for which the GeneratedId was called?

3. How does keytable work. Does it acquire Id inside a transaction. If it does, does this mean that during large updates other threads are blocked because the table with Ids is in transaction / or does this mean that one of two active transactions will fail because the same Ids are being used?

4. Is it possible to change the physical mapping of a field? For example is it possible to change int16 to create a int instead of smallint?

5. There seems to be a bug in Oracle (11g) when using int and using schema synchronization. All ints are automatically converted to Number(38). When lightspeed is checking the structure it sees number(38) and it wants to change to smallint again and again.

6. I noticed that SQL logging does not show anything if there is a syntax error in the SQL. Is it possible to see the SQL even on errors and how?

7. I wrote twice to your contact form and email but haven't received any answer. Do you even offer email support? We are in the process of deciding for a ORM framework and I wanted to ask if it is possible to get a trial to see if model handles 100+ tables well.

1. Yes, but the base class must be derived from Entity<TId> (or a closed constructed type such as Entity<int>).

2. At the moment, yes, overriding GeneratedId is the only way to do this. To get the entity type, call this.GetType() -- we don't currently expose the physical table name but we can do this via the metadata API if required. The field is always Id.

3. Yes, KeyTable acquires each block of IDs inside a transaction. This can result in blocking if multiple threads are trying to get large numbers of IDs in very small batches. To avoid this, use a larger IdentityBlockSize -- then each thread can get a large number of IDs in each transaction, meaning fewer transactions and reduced contention.

4. You mean during designer-database sync? You can't do this directly on the field, but you could probably do it using a user-defined type. And if you are using migrations, you can just change the field type directly in the generated migration starter code.

5. Thanks -- we'll investigate this!

6. SQL logging should show all SQL that gets sent to the database, even if it fails due to a syntax error. Can you post a small sample where logging fails, so we can investigate please? Thanks!

7. One of my colleagues will contact you about this.

1 How would I go about doing this? I would like to override some default behaviour on Entity.

3 Sorry, maybe I did not write it well. Is acquirement of IDs performed in different transaction then the one used for commiting the inserts? Because we had problems with duplicate IDs being acquired when using the same/nested transactions for acquirement of IDs. Just to double check.

4 Yes, I was working with designer. I will check the migrations.

6 I believe it had to do with using table names with spaces without quoting being turned on. I will check and try to reproduce it again.

Thank you for quick answers!

(1) For hand-coded entities, just specify the base class in C# in the normal way e.g. public class Product : EntityBase. For the designer, see http://www.mindscapehq.com/documentation/lightspeed/Working-with-Models-in-the-Visual-Designer/Designer-Shortcuts-and-Tips and scroll down to 'Using a Custom Base Class for Your Entities.'

(3) Yes, it is done in a separate transaction. In fact, the acquisition of IDs is done long before any INSERT -- it occurs when an entity is added to the unit of work, not when the unit of work is flushed to the database. (But even if there is a transaction in place when the entity is added, we still use a separate transaction to allocate IDs.)

(6) My mistake. It does not write out the query immediately. However if you press Run after the exception it does write it out.