Skip to content

deps: [branch-1.1] bump arrow and parquet to 59.3.2 (#7419) - #7448

Merged
andygrove merged 1 commit into
apache:branch-1.1from
fresh-borzoni:backport-7419-branch-1.1
Nov 3, 2026
Merged

andygrove merged 1 commit into
apache:branch-1.1from
fresh-borzoni:backport-7419-branch-1.1

Conversation

@fresh-borzoni

@fresh-borzoni fresh-borzoni commented Nov 3, 2026 •

Copy link
Copy Markdown
Member

Which issue does this PR close?

Backport of #7419 to branch-1.1, for #7417.

Rationale for this change

The native Iceberg scan fails with cannot skip miniblock of size N when position deletes make it skip a long run of rows in a DELTA_BINARY_PACKED column with miniblocks wider than 64 values, which Snowflake and Databricks Photon write (#7417). branch-1.1 locks parquet 59.3.0 and has the bug. arrow-rs 59.3.2 fixes it.

Moving from 59.3.0 also picks up 59.3.1, a patch release of bug fixes that main has had since #7169.

The fix also goes to branch-1.2 in #7447, opened together with this one. Per the backporting guide, this one should not merge before that one is ready. branch-1.0 locks parquet 58.4.0, whose decoder fails the same way, but no 58.x release has the fix (58.4.1 doesn't), so it gets no backport here.

What changes are included in this PR?

git cherry-pick -x 644a870 of #7419, with one adaptation. The lockfiles conflicted, because this branch locks 59.3.0 rather than main's 59.3.1. I kept the branch's lockfiles, reran cargo update -p arrow --precise 59.3.2 and cargo update -p parquet --precise 59.3.2 in native/ and contrib/delta/native/, and restored itertools 0.14.0 for prost-build and prost-derive, as #7419 does. The lockfiles change only the arrow-, parquet and parquet-variant crates. The RSS frame-size test change applied cleanly.

How are these changes tested?

On this branch:

  • ./mvnw test-compile
  • In native/: cargo fmt --all -- --check, cargo check --workspace --all-targets --locked, cargo clippy --workspace --all-targets --locked -- -D warnings and cargo test --workspace --locked (1861 passed)
  • cargo check --locked --all-targets in contrib/delta/native

#7419 adds no Comet test, as arrow-rs covers the decoder. The reproduction in #7417 fails on 1.1.0, and returns Spark's results on 1.2.0-rc1 built with parquet 59.3.2.

* deps: bump arrow and parquet to 59.3.2

arrow-rs 59.3.2 includes apache/arrow-rs#11754, the 59.x backport of
apache/arrow-rs#11021: DeltaBitPackDecoder::skip reads miniblocks wider
than 64 values instead of failing with "cannot skip miniblock of size N".
Snowflake writes 512-value miniblocks and Databricks Photon 256, and a
native scan reaches skip when it skips a long run of rows, such as a run
of Iceberg position deletes.

The native and Delta contrib lockfiles move arrow-*, parquet and
parquet-variant* from 59.3.0 to 59.3.2. See apache#7417.

* deps: keep prost-build and prost-derive on itertools 0.14.0

cargo update re-picked itertools 0.13.0 for them when moving arrow and
parquet. Restore the previous choice, so the lockfiles only change the
arrow and parquet crates.

* test: update the RSS frame size for arrow-rs 59.3.2

(cherry picked from commit 644a870)
@github-actions github-actions Bot added dependencies Pull requests that update a dependency file area:shuffle Shuffle (JVM and native) area:scan Parquet scan / data reading area:Delta Delta Lake contrib (native scan, deletion vectors) labels Nov 3, 2026
@andygrove
andygrove merged commit 17700f1 into apache:branch-1.1 Nov 3, 2026
215 of 217 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

area:Delta Delta Lake contrib (native scan, deletion vectors) area:scan Parquet scan / data reading area:shuffle Shuffle (JVM and native) dependencies Pull requests that update a dependency file

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants