ranflarion opened a new issue, #11109:
URL: https://github.com/apache/arrow-rs/issues/11109
### Describe the bug
`NullBuilder::finish` is documented as "Builds the [NullArray] and reset
this builder", but it never resets `len`; its body is identical to
`finish_cloned`. A `NullBuilder` reused after `finish` therefore emits arrays
whose lengths accumulate across calls, unlike every other builder.
`ArrayBuilder::finish` delegates to the inherent method, so builders obtained
through `make_builder(&DataType::Null, ..)` behave the same way.
### To Reproduce
```rust
use arrow_array::builder::{ArrayBuilder, NullBuilder};
let mut builder = NullBuilder::new();
builder.append_nulls(10);
assert_eq!(builder.finish().len(), 10);
assert_eq!(builder.len(), 0); // fails: left 10
builder.append_nulls(3);
assert_eq!(builder.finish().len(), 3); // fails: left 13
```
### Expected behavior
After `finish`, `len()` is 0 and the next `finish` returns only what was
appended since, matching the doc comment and the other builders.
`finish_cloned` already behaves correctly and should stay non-resetting.
### Additional context
Reproduced on `main` at 8e517524b (after 60.0.0). Found through a shuffle
writer that keeps one builder per output chunk: the `Null` column's emitted
chunks summed to more rows than the partition held, and a downstream length
check caught the mismatch.
--
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]
