[BUG] Iceberg truncate transform does not handle decimal overflow #4016
Description
Describe the bug
It's a bug in Iceberg, this issue is just for tracking Iceberg bug purpose.
Iceberg behavior:
// truncate decimal -999999.999 to -1000000.000, without overflow
BigDecimal in = new BigDecimal("-999999.999");
BigDecimal out = truncateDecimal(new BigInteger("10"), in);
System.out.println(in.precision()); // print 9
System.out.println(out.precision()); // print 10
System.out.println(in); // print -999999.999
System.out.println(out); // print -1000000.000When truncate min value -999999.999 for decimal 32 type, get result value -1000000.000, the result value is type of decimal64.
Decimal 32 supports max precision 9, while Decimal 64 supports max precision 18.
It's obvious that the precision of -1000000.000 is 10, should use decimal 64.
Note: For the integral types, Iceberg just let overflow happen.
Iceberg reports error when write overflow decimails.