xiedeyantu opened a new pull request, #22305:
URL: https://github.com/apache/datafusion/pull/22305
## Which issue does this PR close?
- Closes #22228 .
## Rationale for this change
`array_repeat` still panics for oversized repeat counts in the
constant-scalar path. The simplest reproducer is:
```sql
SELECT array_repeat(1, 9223372036854775807)
```
Unlike the previously reported `array_repeat` overflow cases, this path does
not sum counts across rows and does not multiply nested list lengths, but it
still reaches an unchecked `Vec` preallocation and panics with `capacity
overflow`.
This change makes `array_repeat` reject oversized output lengths up front
and return a normal execution error instead of panicking.
## What changes are included in this PR?
This PR adds explicit bounds checks in repeat.rs so `array_repeat` validates
requested output sizes before allocating buffers.
The main changes are:
- Move repeat-length accumulation into shared checked helpers.
- Reject oversized output lengths with:
`array_repeat: requested length exceeds maximum array size`
- Guard both scalar and list repeat paths so they fail consistently before
hitting unchecked allocation or arithmetic overflow.
- Reuse precomputed outer offsets for the list path instead of rebuilding
them from unchecked lengths.
## Are these changes tested?
Yes.
This PR adds a regression test in repeat.rs covering the constant-scalar
reproducer with `i64::MAX` as the repeat count and verifies that `array_repeat`
returns an execution error rather than panicking.
Validated with:
```bash
cargo test -p datafusion-functions-nested
scalar_count_exceeding_max_array_size_returns_error --lib
```
## Are there any user-facing changes?
Yes.
Previously, oversized `array_repeat` calls could panic the process. After
this change, they return a regular execution error:
```text
array_repeat: requested length exceeds maximum array size
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
