The ORM Foundation

I have created some sample populations, then modified the model such that the populations are not complete. This results in an error such as "Error 2 FactType instance '? has 1.0' in model 'IT Tool Integration' has a partial population." I have not found a simple way to satisfy NORMA, once I get into this mess.

Is this 'error' really fatal - in that the model is inconsistent and could not generate SQL?  Or should this really be notied as a warning?

Jim, 

This error supports automatic activation. To activate the sample population editor to fill out the population:

• From the Error List, double click the error to select the FactType, open the Sample Population Editor, and provide an instance for the blank cell(s).
• In the verbalization browser, the error in question will display with the FactType verbalization and is hyperlinked to activate the error.
• From the FactType shape on diagram, double-click the edge of the FactType to activate the next error.

If you don't want to see population errors, then

• Click on the backgroun of an ORM diagram
• Open the Properties Window and activate the 'Error Display' dialog
• Uncheck the 'Sample Population Errors' category of errors, or expand and turn off the specific 'Incomplete FactType Sample Population' error

These errors have no affect on SQL generation. In fact, SQL generation is done for any FactType and ObjectType that is sufficiently error free. For example, a FactType with no internal uniqueness or a ValueType with no DataType does not map, but these elements are mapped as soon as the error states are cleared. Sample population errors are not critical for SQL generation.

Warning vs Error is a matter of perspective. If your goal is to provide a fully validated model with no population errors, then this is definitely an error. If you want SQL structure, then it is probably a warning. If you want to prepopulate the SQL model [we don't do this yet], then it is an error again. Short of controlling error display, NORMA doesn't currently have a mechanism for error level. This could possibly be added in the future, along with finer-grained control over error display (display graphically, on model browser, in verbalization browser and reports, and in the error list). However, these additional facilities are not high on our priority list.

-Matt

Thanks Matt!

The instruction on adjusting the error display is very helpful. At least that puts them out of my mind.

It might still be helpful to have a way to clear out a sample population that has become so outdated (incomplete) that it is more of a hindrance than a help.

Jim,

There's something funky going on with the Delete in the Sample Population editor. You should always be able to Delete a row by selecting the first column (the one with the row number in it) and pressing Delete. However, it looks like the delete command is disabled after escaping out of an inline edit.

In the meantime, this should always work:

1. Select the FactType you want to empty the population for in a diagram (or the Model Browser) while the Sample Population tool window is open.
2. Click in the first (numbered) column of the sample population editor on the instance you want to delete.
3. Press 'Delete'

Note that deleting all entries in the row will also delete the row. Selecting the role player object types and deleting their populations will also indirectly clear the fact instances.

Sorry about the delete problem, I'll look at what's up with the delete command status.

-Matt

Interesting results:

Since I have hidden the sample population errors, I go back to turn these error messages visble. The ErrorDisplay dialog indicates that they should be visible. But no error appears!

So I open the Sample Population Editor window and work through all the Fact Types. None has any sample data now. This is what I wanted, but not the way I expected to get there!

Generally, the main thing that will empty sample populations is changing the preferred identifier on an entity. If the entity instance list is cleared, and an instance is the only instance used in a fact instance, then the fact instance will clear as well.

In any case, the delete behavior was fixed in changeset 1388 (row deletion activates correctly, and deleting an inactive cell corresponds to choose <Unspecified> from an active drop down). The changes are in the May 2009 CTP.

Thanks for the heads up. I've heard this issue mentioned by others since you posted, but you were the first to bring the usability issue to my attention.

-Matt