[SPARK-58235][SQL] Fix Double/Float to Timestamp silent overflow clamping in non-ANSI mode

[jiangxt2]

What changes were proposed in this pull request?

Add Long range validation for the Double/Float-to-Timestamp cast path in both interpreted and codegen execution.

When a Double or Float value multiplied by MICROS_PER_SECOND (1,000,000) exceeds Long range, the unchecked cast to long silently produces a clamped value (Long.MAX_VALUE or Long.MIN_VALUE). The fix detects this overflow and either returns NULL (non-ANSI mode) or throws CAST_OVERFLOW (ANSI mode).

The interpreted path (Cast.doubleToTimestamp) now checks the multiplication result against Long range before calling .toLong. The codegen path (castToTimestampCode Double/Float branches) applies the same range check before the (long) cast.

Why are the changes needed?

CAST(1e20 AS TIMESTAMP) silently produces +294247-01-10 04:00:54.775807 (clamped to Long.MAX_VALUE) instead of NULL in non-ANSI mode. This is inconsistent with other overflow behavior:

SET spark.sql.ansi.enabled=false;
SELECT CAST(1e20 AS TIMESTAMP);
-- Before fix: +294247-01-10 04:00:54.775807  (clamped to Long.MAX_VALUE)
-- After fix:  NULL

SELECT CAST(-1e20 AS TIMESTAMP);
-- Before fix: -290308-12-21 19:59:05.224192  (clamped to Long.MIN_VALUE)
-- After fix:  NULL

The ANSI mode path (doubleToTimestampAnsi → DoubleExactNumeric.toLong) already correctly throws CAST_OVERFLOW. This bug only affects non-ANSI mode.

Does this PR introduce any user-facing change?

Yes. Previously, CAST(large_double AS TIMESTAMP) silently returned an incorrect value caused by numeric overflow. Now:

• Non-ANSI mode: returns NULL (consistent with other overflow casts)
• ANSI mode: throws CAST_OVERFLOW error (unchanged)

How was this patch tested?

Added unit tests in CastWithAnsiOffSuite and CastWithAnsiOnSuite:

Non-ANSI mode (CastWithAnsiOffSuite):

• Positive overflow: 1E20 → null
• Negative overflow: -1E20 → null
• Boundary just under: 9223372036854.0 → passes through
• Boundary just over: 9223372036855.0 → null
• Negative boundary: -9223372036854.0 / -9223372036855.0
• Finite input overflowing to +-Infinity: Double.MaxValue / -Double.MaxValue
• NaN, Infinity, normal values, zero, Float overflow

ANSI mode (CastWithAnsiOnSuite):

• Positive overflow: throws CAST_OVERFLOW with correct error parameters
• Negative overflow: throws CAST_OVERFLOW with correct error parameters
• Boundary values, normal values, Float overflow

Was this patch authored or co-authored using generative AI tooling?

Generated-by: Claude Code (Opus 4.8)

[uros-b]

Thank you @jiangxt2, please address comment before merge - otherwise LGTM! cc @MaxGekk @gengliangwang

[jiangxt2]

Thanks for catching this! Moved the entry to the Upgrading from Spark SQL 4.2 to 4.3 section and corrected the version to Since Spark 4.3. @uros-b