apache / apache/arrow-java

[JAVA] FlightSQL JDBC Driver PreparedStatement#setTimestamp() ignores connection timeStampPrecision

Abierto
#838 1 comentario 0 reacciones 0 asignados Ver en GitHub
Type: bug
Lenguaje dominante
Java
Estrellas
94
Forks
152
Merge medio
3 d 16 h
PR fusionados (30 d)
11

Descripción

The FlightSQL JDBC driver always treats timestamps as milliseconds when sending them to the server, ignoring the connection setting timeStampPrecision (for example microseconds). As a result, when the connection precision is set to microseconds the driver truncates/offsets the timestamp and the value stored on the server is incorrect.

Expected behavior When the JDBC connection option timeStampPrecision is set (e.g. microseconds, milliseconds, nanoseconds), PreparedStatement#setTimestamp (and setTimestamp with timestamptz) should send timestamps with the configured precision so the server receives the correct epoch value.

Actual behavior The driver converts java.sql.Timestamp values to an epoch value assuming milliseconds. If the connection is configured for microseconds (or other precision) the conversion is incorrect and inserted timestamps are wrong.

```
String jdbcUrlFlight = "jdbc:arrow-flight-sql://localhost:9994?useEncryption=0";
try (Connection conn = DriverManager.getConnection(jdbcUrlFlight);
PreparedStatement stmt = conn.prepareStatement("INSERT INTO ... VALUES (?, ?, ?, ?, ?)")) {

for (RowData row : generatedData) {
stmt.setDate(1, row.c_date());
stmt.setTime(2, row.c_time());
stmt.setTime(3, row.c_timetz());
stmt.setTimestamp(4, row.c_timestamp()); // incorrect conversion happens here
stmt.setTimestamp(5, row.c_timestamptz()); // and here
stmt.addBatch();
}
stmt.executeBatch();
}
```

Example A — timestamp precision in microseconds (bug triggered)

- Value I want to insert:
- Timestamp object: "2024-11-03 12:45:09.869885001"
- nanos = 869885001
- fastTime = 1730634309000

- **DateTimeUtils#sqlTimestampToUnixTimestamp(Timestamp, TimeZone) returns: 1730637909869**:
- Interpreted as milliseconds, that is: **2024-11-03 12:45:09.869** (GMT)

- Final value inserted into server: **1970-01-21 00:43:57.909869**
- Incorrect because the connection used microsecond precision but driver treated the value as milliseconds.

Example B — timestamp precision in milliseconds (works)

- Value I want to insert:
- "**2023-11-29 09:47:32.659534860**"
- nanos = 659534860
- fastTime = 1701247652000
- Driver sends: 1701251252659 (interpreted as milliseconds)

- Server receives: **2023-11-29 09:47:32.659** — correct when precision is milliseconds.

Environment
- flight-sql-jdbc-driver-18.3.0.jar
- Java: 17
- Server FlightSQL implementation/version: 18.2
- OS: Windows

Relevant code & suspected area

DateTimeUtils#sqlTimestampToUnixTimestamp(Timestamp, TimeZone) appears to be converting timestamps assuming millisecond precision.
```
public static long sqlTimestampToUnixTimestamp(Timestamp timestamp, TimeZone timeZone) {
long time = timestamp.getTime();
LocalDateTime dateTime = timestamp.toLocalDateTime();
long unixTimestamp = dateTime.toEpochSecond(ZoneOffset.UTC) * 1000L + (long)dateTime.get(ChronoField.MILLI_OF_SECOND);
if (timeZone != null) {
unixTimestamp += (long)timeZone.getOffset(time);
}

unixTimestamp -= (long)DEFAULT_ZONE.getOffset(time);
return unixTimestamp;
}
```

The driver should respect the connection timeStampPrecision parameter, and correctlyreturn the exact epoch value (as a long) at the configured timestamp precision.

Guía de contribución

Abrir la guía de contribución

Línea de trabajo

Start by tracing PreparedStatement#setTimestamp and setTimestamp for timestamptz into DateTimeUtils#sqlTimestampToUnixTimestamp(Timestamp, TimeZone), then inspect how the connection timeStampPrecision option is exposed to that path. Reproduce the microsecond and millisecond examples from the issue and add or update focused coverage if the relevant test location is identified. Done means the server receives the correct epoch value for each configured precision.

Escrito por el modelo de indexación a partir del texto del issue.

Evaluación

Stack tecnológico
java
Área
databases
Tipo de issue
Error
Dificultad
3/5
Tiempo estimado
1-2 días
Estado de actividad
Estancado
Claridad
Bien especificado
Aptitud para principiantes
48/100

Recibe los nuevos issues en tu correo

Un resumen breve de issues de GitHub para principiantes.