Using PostgreSQL, I get wrong result due to large numeric values on dimensions being rounded up. It seems that Mondrian (also internally) formats the number as scientific E notation, but with too low precision.
I create and populate the following table:
Next I make the following schema:
Then I run the following MDX:
The measure is empy, not 1 as expected:
Looking at the log, it seems that Mondrian fetches the dimension value (9123372036854775807), rounds it up to 9.1233720368547758E18, and then retrieves all rows where foo equals the rounded up value, not the exact value. Therefore no rows are found.
I create and populate the following table:
Code:
create table public.test (foo bigint, bar int);
insert into public.test values (9123372036854775807,1);Code:
<Schema name="Test">
<Dimension name="Foo">
<Hierarchy hasAll="true">
<Table name="test" schema="public"/>
<Level name="Foo" column="foo" type="Numeric">
</Level>
</Hierarchy>
</Dimension>
<Cube name="Test">
<Table name="test" schema="public"/>
<DimensionUsage name="Foo" source="Foo"/>
<Measure name="Bar" column="bar" aggregator="sum"/>
</Cube>
</Schema>Code:
SELECT
[Foo].[Foo].MEMBERS ON COLUMNS
,[Measures].[Bar] ON ROWS
FROM [Test]| [Foo].[9.1233720368547758E18] | |
| [Measures].[Bar] |
Looking at the log, it seems that Mondrian fetches the dimension value (9123372036854775807), rounds it up to 9.1233720368547758E18, and then retrieves all rows where foo equals the rounded up value, not the exact value. Therefore no rows are found.