Stefan S. created CAY-2623:
------------------------------
Summary: Directive #result: Column descriptor mismatch leads to
wrong data row keys and missing mappings
Key: CAY-2623
URL: https://issues.apache.org/jira/browse/CAY-2623
Project: Cayenne
Issue Type: Bug
Affects Versions: 4.1.B2
Reporter: Stefan S.
According to current documentation, the variant of #result "#result(column
javaType alias dataRowKey)" is intended for mapping resulting fields of queries
to complex DB path expressions which cannot be modeled as alias of fields due
to SQL syntax violations.
Unfortunately such fields are not mapped, because the key within the
corresponding DataRow field is not equal to 'dataRowKey'.
Debugging org.apache.cayenne.access.jdbc.RowDescriptorBuilder and
org.apache.cayenne.template.directive.Result revealed that the corresponding
ColumnDescriptor is not found.
The key issue seems to be in RowDescriptorBuilder.getColumentDescriptor method,
in the following line:
if (columnRowKey != null && columnRowKey.equalsIgnoreCase(rowKey)) {
return columnArray[i];
}
columnRowKey is actually the data row key BUT rowKey is either the alias or the
field name of the column. Per definition for the intended case (see above),
they are never equal and hence the ColumnDescriptor is not applied.
Changing one of the attributes of the condition with the debugger in a way it
returns the right ColumnDescriptor, the eventual resulting data object are
correct.
I'd propose to add (or replace) a comparison of ColumnDescriptor and Column by
column alias and / or column name. I'd prefer using the alias (if available)
first, since this is IMHO the purpose of an alias. However, that would require
to extend ColumnDescriptor class to hold also the alias of the column and to
extend Result to apply the alias value to the ColumnDescriptor.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
