[VibeSWE]

Problem

extract_timestamp_bound (src/query/stream_schema_provider.rs) called DateTime::from_timestamp_millis(v).unwrap(). Any SQL query whose millisecond timestamp literal is outside chrono's range panics the query worker during planning:

SELECT * FROM logs WHERE p_timestamp >= to_timestamp_millis(9223372036854775807)

The panic happens in supports_filters_pushdown → expr_in_boundary, so it is reachable on every stream regardless of the manifest-pruning column checks.

Found while reviewing #63 (a GPT-6.1 Sol review subagent reproduced it); independent of that PR.

Fix

Return None (no time bound) when the literal can't be represented. The filter is then reported Inexact, contributes no manifest-pruning bound, and DataFusion evaluates it normally. The nanosecond arm already uses chrono's infallible from_timestamp_nanos.

Tests

• out_of_range_millisecond_literal_yields_no_bound: unit test on extract_timestamp_bound / expr_in_boundary.
• out_of_range_millisecond_literal_in_sql_does_not_panic: plans the SQL above through a SessionContext.

Both panic on main and pass with the fix. cargo fmt --check is clean, cargo test passes (851 passed), and cargo clippy --all-targets reports nothing new in the changed file.

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}