BigQuery: FieldValue.getTimestamp() use of Double causes precision loss in some cases.

[slilichenko]

Environment details

1. Specify the API at the beginning of the title (for example, "BigQuery: ...")
   General, Core, and Other are also allowed as types
2. OS type and version: any
3. Java version: any
4. google-cloud-java version(s): 1.96.0

Steps to reproduce

Create a row with this timestamp: 1337-09-02 11:43:21.622894 UTC, query and use fieldValue.getTimestamp(). Microseconds wouldn't be equal.

Code example

@test
public void testFloatingPointPrecisionLoss() {
FieldValue fieldValue = FieldValue.of(Attribute.PRIMITIVE, "-1.9954383398377106E10");
long received = fieldValue.getTimestampValue();
long expected = -19954383398377106L;
assertEquals(expected, received);
}

Any additional information below

Cause - floating point conversation causes it; need to use BigDecimal (or equivalent) to perform lossless conversion.

[slilichenko]

Sorry, but it's not a question, it's a bug...

[Obstinaty]

退订

[shollyman]

This is a bug that necessitates changes in the API backend. There's work in progress to address this.

[slilichenko]

I think it's much simpler than that: in FieldValue.getTimestamp() replace
new Double(Double.valueOf(getStringValue()) * MICROSECONDS).longValue()
with:
BigDecimal secondsWithMicro = new BigDecimal(getStringValue());
BigDecimal scaled = secondsWithMicro.scaleByPowerOfTen(6);
return scaled.longValue();
Obviously, can be made faster with a bit of parsing by hand rather than using BigDecimal.