JunRuiLee opened a new pull request, #751:
URL: https://github.com/apache/paimon-rust/pull/751
## What
`IndexFileMeta._ROW_COUNT` is an Avro `long` in Java (`IndexFileMeta.SCHEMA`
field 3,
`new DataField(3, "_ROW_COUNT", new BigIntType(false))`), but the Rust field
is `i32` and the decoder
narrows it:
```rust
"_ROW_COUNT" => row_count = Some(read_long_field(cursor, field.nullable)? as
i32),
```
An index file with more than `i32::MAX` rows therefore decodes to a silently
wrong (possibly negative)
row count. The index manifest is written by Java and read here, so nothing
surfaces the truncation.
Widen the field to `i64` and drop the cast. The substantive change is three
lines
(`IndexFileMeta.row_count`, the decoder's local, and the cast); everything
else is mechanical
adaptation of construction sites and literals across the crate, the
fulltext-gated code, the tests and
the C bindings.
## Assumptions and limits, called out for reviewers
- This is a latent-overflow fix, reachable only above `i32::MAX` rows in a
single index file. It is not
triggered by any existing fixture, so the new test constructs the case
directly.
- Widening a public field is a source-compatibility change for downstream
code that constructs
`IndexFileMeta` or reads `row_count` with a typed binding.
## Testing
`row_count_above_i32_max_round_trips_through_index_manifest` fails before
the change and passes after.
Gates run locally from a clean HEAD: `cargo fmt --all -- --check`; `cargo
clippy -p paimon
--all-targets -- -D warnings` and the same with `--features fulltext`, both
clean; `cargo test -p
paimon --lib` 2406 passed / 0 failed and 2480 passed / 0 failed with
`--features fulltext`.
## Note
Opened as a draft. This is the first of two independent index-manifest
compatibility fixes; the second
one (index file path resolution) is stacked on top of this branch and is
sent separately. There is no
semantic dependency between them — they share ~19 files of struct-literal
sites, which is why they are
ordered rather than concurrent.
--
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]
