We're seeing Geodatabase.ApplyEdits() leaving what appear to be unnecessary records in the SDE.STATES and SDE.STATE_LINEAGES tables when we edit non-SDE Versioned data, and are wondering what the cause is.
This from within a "headless" console application, which is why we are using Geodatabase.ApplyEdits(). We do not see this issue when using EditOperation.ExecuteAsync, from within an ArcGIS Pro add-in.
If we point the code below at a non-SDE Versioned Feature Class, we end up with additional records in the SDE.STATES and SDE.STATE_LINEAGES tables.
For example - if we have only one row in both SDE.STATES and SDE.STATE_LINEAGES, and we run the code below that inserts a row into a non-versioned feature class, at the end of the run we'll have two records in SDE.STATES and three records in SDE.STATE_LINEAGES.
If there are lots of updates to a non-versioned feature class in a day, that means our STATES and STATE_LINEAGES tables grow by quite a bit. Which means the user has to compress the database more frequently and/or takes a performance hit on querying versioned layers in the database.
Here's the code we are executing (it's based on the ProConcepts Geodatabase ApplyEdits sample😞
DatabaseConnectionProperties connectionProperties = new DatabaseConnectionProperties(EnterpriseDatabaseType.Oracle)
{
AuthenticationMode = AuthenticationMode.DBMS,
Instance = @OUR_SERVER,
User = "xyz",
Password = "xyz",
Version = "SDE.DEFAULT" // We've also tried setting this to NULL and not setting it at all.
};
using (Geodatabase gdb = new Geodatabase(DatabaseConnectionProperties))
{
using (FeatureClass featureClass = gdb.OpenDataset<FeatureClass>("TEST_NONVERSION_FC"))
{
FeatureClassDefinition fcDefinition = fc.GetDefinition();
gdb.ApplyEdits(() => {
using (RowBuffer rowbuffer = featureClass.CreateRowBuffer())
{
// set an attribute value
rowbuffer[1] = 12;
// create a geometry
List<MapPoint> points = new List<MapPoint>();
points.Add(MapPointBuilder.CreateMapPoint(-85.003, 47.012800, 100, 0, fcDefinition.GetSpatialReference()));
points.Add(MapPointBuilder.CreateMapPoint(-105.994, 26.368, 1000, 1000, fcDefinition.GetSpatialReference()));
using (PolylineBuilder pb = new PolylineBuilder(points))
{
pb.SpatialReference = fcDefinition.GetSpatialReference();
pb.HasM = true;
pb.HasZ = true;
rowbuffer[fcDefinition.GetShapeField()] = pb.ToGeometry();
}
// insert the data into the feature class
using (Feature f = featureClass.CreateRow(rowbuffer))
{ }
}
});
}
}
Is the above code's workflow more or less correct?
Is there an additional cleanup we need to perform?
