[Mondrian] [Fwd: Eigenbase perforce change 9873 for review]

John V. Sichi jsichi at gmail.com
Wed Sep 12 02:51:13 EDT 2007


Julian Hyde wrote:
> The dubious thing in Rushan's change is the assumption that a Double value
> in java maps to a DOUBLE column in SQL. The datatype of the value in java
> depends as much on the JDBC driver as anything. I guess the driver should
> use BigDecimal for fixed-point values, but many of them use Double instead.
> 
> I wouldn't think it is safe to extend this assumption to other databases and
> JDBC drivers, so the prudent thing is to keep this as a LucidDB-specific
> change.

OK.  But keep in mind that if the value is fixed-point in the DBMS, yet 
gets returned as Double via JDBC, then there's already a potential 
problem due to loss of precision, e.g.

1.234567890123456789

can't be represented as a double, so after it gets rounded during fetch, 
and then Mondrian sends it back to the DBMS as a literal like 
1.234567890123457, the search predicate won't match.

JVS



More information about the Mondrian mailing list