{ NUMBER [ (precision [, scale ]) ]
| FLOAT [ (precision) ]
| BINARY_FLOAT
| BINARY_DOUBLE
}