Postgres, for example, allows NaN and infinity values on numeric columns (with
some exclusions).
When we try read these, we get:
org.apache.metamodel.MetaModelException: Could not get next record in
resultset: Bad value for type BigDecimal : NaN
at org.apache.metamodel.jdbc.JdbcUtils.wrapException(JdbcUtils.java:61)
at org.apache.metamodel.jdbc.JdbcDataSet.next(JdbcDataSet.java:141)
. . .
Caused by: org.postgresql.util.PSQLException: Bad value for type BigDecimal :
NaN
at org.postgresql.jdbc.PgResultSet.toBigDecimal(PgResultSet.java:2880)
at org.postgresql.jdbc.PgResultSet.toBigDecimal(PgResultSet.java:2889)
at org.postgresql.jdbc.PgResultSet.getBigDecimal(PgResultSet.java:2332)
at
org.postgresql.jdbc.PgResultSet.internalGetObject(PgResultSet.java:192)
at org.postgresql.jdbc.PgResultSet.getObject(PgResultSet.java:2571)
at
org.apache.metamodel.jdbc.dialects.AbstractQueryRewriter.getResultSetValue(AbstractQueryRewriter.java:461)
In our previous code, we would try these as BigDecimal, and if a SQL exception
occurred, we would retry it with ResultSet.getDouble() which works because
Double supports both Nan and Infinite values.
Can you suggest a work-around or fix? We have thought of a couple, but because
database vendors differ in their support for NaN and Infinite values, none of
them are very appealing.
Best,
Dana
