Push expressions through layout plans

[joseph-isaacs]

Summary

• replace the virtual Plan::optimize_expression hook with static typed parent-child rules
• register rules for (ExpressionPlan, StructPlan), (ExpressionPlan, DictPlan), (ExpressionPlan, ChunkedPlan), and (ExpressionPlan, RowIdxPlan)
• push independent struct-field expressions into each field while retaining cross-field work as a residual expression
• push safe boolean expressions into dictionary values while retaining dictionary codes
• optimize heterogeneous chunks independently while preserving global row-index semantics
• partition mixed row-index/data expressions and combine their outputs with a residual expression
• add complete before/after snapshots, including multi-field struct and row-index/struct composition

Stack

This draft is stacked on #9196 and targets vortex-plan-rules.

Example

Before optimization:

ExpressionPlan((#row_idx > 11) and (($.a > 5) and ($.b > 7)))
  RowIdxPlan
    StructPlan
      a: DictPlan
      b: FlatPlan

After optimization, the row-index predicate is evaluated by RowIdxValuesPlan, the field predicates are pushed into the corresponding struct children, and a residual expression combines the partition outputs. No synthetic struct-partition plan node is introduced.

Safety

• dictionary pushdown requires boolean output, a root reference, strictness, and infallibility
• nullable struct expressions remain above parent validity
• chunk pushdown rejects expressions referencing global row indices
• parent-reduction rules verify row-count and dtype preservation in debug builds
• cross-field expressions that cannot be partitioned remain above the original struct plan

Checks

• cargo test -p vortex-layout — 209 passed
• cargo clippy -p vortex-layout --all-targets --all-features -- -D warnings
• cargo +nightly fmt --all -- --check
• git diff --check

[codspeed-hq]

Hooray! CodSpeed harness just leveled up!

The base and head of this comparison were measured with different runner settings, so their benchmark values are not directly comparable.

What changed between base and head:

• CodSpeed runner v5 changed how benchmarks are measured (base 5.0.1 → head 4.13.0). View release notes
• Cycle estimation (on → off) changes how benchmark performance is calculated. Learn more about cycle estimation

Re-run the base with the same settings to get a valid performance comparison.

---

_{Comparing vortex-plan-optimizer (a9043d5) with develop (3884b98)¹}

Footnotes

1. No successful run was found on vortex-plan-rules (ed09b04) during the generation of this report, so develop (3884b98) was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩