refactor(proto): migrate scalar subquery serde - #23915
Merged
Merged
Conversation
ScalarSubqueryExpr nodes must share the results container populated by their enclosing ScalarSubqueryExec. A normal child decode loses that scope when serialization moves into the plan implementation. Add a scoped child decode operation so the new plan hooks preserve the shared container while removing the central serialization path. Closes apache#23515 Signed-off-by: Jiawei Zhao <Phoenix500526@163.com>
|
Thank you for opening this pull request! Reviewer note: cargo-semver-checks reported the current version number is not SemVer-compatible with the changes in this pull request (compared against the base branch). Details |
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #23915 +/- ##
==========================================
- Coverage 80.66% 80.65% -0.02%
==========================================
Files 1095 1095
Lines 372294 372319 +25
Branches 372294 372319 +25
==========================================
- Hits 300324 300296 -28
- Misses 54055 54101 +46
- Partials 17915 17922 +7 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
adriangb
approved these changes
Jul 30, 2026
|
|
||
| /// Deserialize a child plan with `results` active for scalar subquery | ||
| /// expressions in that plan's subtree. | ||
| fn decode_plan_with_scalar_subquery_results( |
Contributor
There was a problem hiding this comment.
It'd be nice to seal this trait. I'll make a PR.
Contributor
There was a problem hiding this comment.
We can't seal it, but I opened #24001. Should not block this PR.
adriangb
enabled auto-merge
July 30, 2026 14:30
rluvaton
pushed a commit
to rluvaton/datafusion
that referenced
this pull request
Jul 30, 2026
…API (apache#24001) ## Which issue does this PR close? - Part of apache#23494. Precursor to apache#23915 (and to the remaining `DataSource` / `DataSink` items, apache#23497 / apache#23498). ## Rationale for this change `ExecutionPlanEncode` / `ExecutionPlanDecode` (added in apache#23495) are dispatch details of the `try_to_proto` / `try_from_proto` hooks, not extension points. They are defined in `datafusion-physical-plan`, implemented only by `ConverterPlanEncoder` / `ConverterPlanDecoder` in `datafusion-proto`, and plans reach them exclusively through `ExecutionPlanEncodeCtx` / `ExecutionPlanDecodeCtx` — their own docs already say so: > Internal dispatch trait backing [`ExecutionPlanEncodeCtx`]. Implemented by `datafusion-proto`. > Plan authors never name this trait. They are `pub` only because those adapters live in another crate. Nothing said that to the tooling, so every capability the epic still has to add to the ctx reads as a major breaking change. apache#23915 hits this first: it needs `decode_plan_with_scalar_subquery_results` so `ScalarSubqueryExec` can decode its input with the subquery-results container in scope, and `cargo-semver-checks` flags the required method as `trait_method_added`. The `DataSource` / `DataSink` families will want their own primitives next. `#[doc(hidden)]` states what was already true, and makes those additions changes to something that was never public API — rather than asking each follow-up PR to explain away a breakage report. ## What changes are included in this PR? Mark `ExecutionPlanEncode` and `ExecutionPlanDecode` `#[doc(hidden)]`, and say why in their docs and in the module overview. One file, no behavior, wire-format, or signature changes. A sealed-trait supertrait was the first thing I tried. It is worse here: sealing across a crate boundary needs a `pub` marker anyway, so it does not actually prevent a downstream impl — it just adds a public item in order to say "this is not public API", plus an `impl` line for every implementor including test doubles. `#[doc(hidden)]` says the same thing by removing API surface instead of adding it, and `cargo-semver-checks` honors both identically (measured below). ## Are these changes tested? The property this PR buys is a `cargo-semver-checks` classification, so it is verified with that tool directly — v0.49.0, the version CI installs, invoked the way CI invokes it: | baseline | change under test | result | |---|---|---| | `main` | add a required method to `ExecutionPlanDecode` | `trait_method_added` — **major** | | this PR | the same required method | 223 checks pass, **no semver update required** | Worth recording, since it drove the shape of this PR: the lint reads `public_api_sealed` from the **baseline**, so this has to land before the PRs that add methods, not alongside them. Also, `cargo-semver-checks` does not re-qualify the traits as public API even though `ExecutionPlanEncodeCtx::new` still names them in a public signature. Also ran, on the pinned 1.97.0 toolchain: `cargo fmt --all -- --check`, `cargo clippy --all-targets --all-features` for both crates, `cargo doc` with `-D warnings` (no broken intra-doc links from the module docs to the now-hidden traits), and the full `datafusion-proto` integration suite (211 passed). ## Are there any user-facing changes? `cargo-semver-checks` will report `trait_now_doc_hidden` on this PR, and that is the intended content of the change. Real-world impact is nil: both traits were introduced by apache#23495, which merged after `54.1.0` was tagged, so no released version of `datafusion-physical-plan` contains them. The load-bearing public API — `ExecutionPlan::try_to_proto`, the two ctx types, `AsExecutionPlan`, `PhysicalExtensionCodec`, `PhysicalProtoConverterExtension` — is untouched, and `datafusion-proto` needs no change at all. Note for contributors on apache#23494: the expression-side equivalents (`PhysicalExprEncode` / `PhysicalExprDecode` in `physical-expr-common`) have the same shape and the same argument, but several test doubles across crates implement them. Left for a follow-up. Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Which issue does this PR close?
Rationale for this change
ScalarSubqueryExpr nodes must share the results container populated by their enclosing ScalarSubqueryExec. A normal child decode loses that scope when serialization moves into the plan implementation.
Add a scoped child decode operation so the new plan hooks preserve the shared container while removing the central serialization path.
What changes are included in this PR?
Are these changes tested?
Are there any user-facing changes?