linq_fold: PR D1 — group_by adapter reconciliation + 2 migrations + reducer-spec table

[borisbat]

Summary

Continues the linq_fold pattern-table refactor (PRs A / B1 / B2 / C merged). PR D1 is the first half of PR D's masterplan scope: tackle the group_by family. Pure refactor — semantics unchanged.

• Bundled R1 follow-up to linq_fold: PR C — SourceAdapter + 4 decs planner migrations #2885: drop dead retType in emit_bounded_heap (3 LOC). Return type was already being inferred; the assignment was dead.
• Partial GroupBy ↔ SourceAdapter reconciliation: new to_source_adapter(GroupBySourceAdapter) : SourceAdapter projection. adapter_emit_source_loop keeps decs-join inline (~40 LOC, no SourceAdapter variant for it yet) but delegates array + plain-decs to PR C's adapter_wrap_source_loop. adapter_finalize_emission collapses to a one-liner over adapter_wrap_invoke. Full unification deferred to PR D2 (when plan_zip / plan_decs_join migrate and Zip/DecsJoin variants join).
• plan_group_by + plan_decs_group_by → pattern-table stubs: 2 SplicePattern rows (group_by_array, group_by_decs) + 1 shared emit_group_by delegating to plan_group_by_core (unchanged sub-codegen). Slot.arity enforces the args-length guards (count=1, join=5, group_by_lazy=2, order_by[descending]=2, where=2). New order_by_family alias excludes bare order / order_descending (no key arg). New decs_join_invariants predicate enforces v1 join-shape limits (empty head, no having, no trailing_where when upstream_join captured). ~165 LOC of imperative tail-pop hard-deleted across both planners.
• emit_reducer_branches 13-arm if/elif → reducer_emitters data table: table<string; ReducerEmitterFn> lookup with 10 named mk_reducer_* fns. Shared mk_strictly_preferred(workhorse, isMin, valExpr, cmpExpr) helper collapses the 4-arm workhorse split (workhorse × isMin) shared between mk_reducer_min_max and mk_reducer_min_max_inner. mk_* naming marks these as sub-codegen building blocks, not pattern-table emit fns.

Architectural decisions

• D1-A — Partial reconciliation only. GroupBySourceAdapter stays load-bearing (per-mode field bags diverge; plan_group_by_core reads initialBindName / initialElemType directly per mode). Projection unlocks helper reuse without restructuring.
• D1-B — decs-join → Decs(null, "djoin_jres") is safe at finalize: adapter_wrap_invoke's Decs branch reads only the variant tag, never the tuple fields. Array path's invoke shifts untyped→typed; plan_group_by_core unconditionally sets retType so the inference shift is a no-op.
• D1-C — Decs+join fits as a SINGLE row with optional upstream_join slot (not 2 separate rows). Predicate-gated invariants moved from emit-fn-land to slot/predicate-land via Slot.arity + decs_join_invariants.
• D1-D — Reducer table LOC delta is ~+25 net (helper savings less than the agent's ~30 estimate); win is uniformity + lint-ability + future extensibility, not bytes.

Full decision log in daslib/linq_fold.md (7 new entries: D1-A through D1-E + 2 impl notes).

Out of scope (deferred to PR D2)

• plan_zip (352 LOC) + plan_decs_join (189 LOC) migrations
• SourceAdapter widening to Array | Decs | DecsJoin | Zip
• Full GroupBySourceAdapter unification (once DecsJoin variant exists)
• reverse |> distinct[_by] on decs sources (D6 from PR C plan)

Files

• daslib/linq_fold.das — all implementation
• daslib/linq_fold.md — masterplan refresh (PR D row split into D1 complete + D2 deferred; 7 new decision log entries)
• doc/source/reference/linq_fold_patterns.rst — group_by rows rewritten as pattern <name> (sub-codegen plan_group_by_core) handles …
• benchmarks/sql/results.md — INTERP + JIT matrix refresh for 16 group_by + join_groupby benches

Test plan

• mcp__daslang__compile_check daslib/linq_fold.das — clean
• mcp__daslang__lint daslib/linq_fold.das — clean
• tests/linq/test_linq_group_by.das — 18 passed
• tests/linq/test_linq_fold_theme3_c2_group_by_order_by.das — 14 passed
• tests/linq/test_linq_fold_theme3_decs_join_groupby.das — 14 passed
• tests/linq/test_linq_fold_theme2_trailing_where.das — 22 passed (HAVING shapes)
• tests/linq/test_linq_fold_theme45_quick_wins.das — 32 passed
• tests/linq/test_linq_from_decs.das — 198 passed
• tests/linq/test_linq_fold_ast.das — 228 passed
• tests/linq/test_linq_fold_pattern_walker.das — 16 passed
• tests/linq/test_linq_fold_terminal_select.das — 28 passed
• tests/linq/test_linq_fold_theme7_chained_select.das — 18 passed
• tests/linq/test_linq_fold_theme8_fusion_arms.das — 36 passed
• 624 tests pass across 13 affected files
• INTERP + JIT bench matrix for 16 group_by + join_groupby benches refreshed — all deltas within run-to-run noise (~±1 ns INTERP, ~±0.3 ns JIT)
• Full CI matrix on push (Linux / macOS / Windows, INTERP + AOT + JIT, modulo the 27 documented master JIT-quote regressions)

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR continues the linq_fold pattern-table refactor by migrating the group_by planner family onto pattern-table rows, partially reconciling GroupBySourceAdapter with the newer SourceAdapter helpers, and converting reducer emission dispatch into a data-table lookup. It also updates the corresponding documentation and refreshes benchmark result matrices for the affected group-by benchmarks.

Changes:

• Migrate plan_group_by / plan_decs_group_by to pattern-table stubs backed by shared emit_group_by → plan_group_by_core.
• Introduce to_source_adapter(GroupBySourceAdapter) and reuse adapter_wrap_source_loop / adapter_wrap_invoke for non-join group-by sources.
• Replace emit_reducer_branches if/elif chain with reducer_emitters table + mk_reducer_* helpers; update docs/bench outputs accordingly.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
daslib/linq_fold.das	Core refactor: group_by patterns/emitter, adapter projection, reducer emitter table, and minor cleanup in bounded-heap emission.
daslib/linq_fold.md	Masterplan/status + decision log updates documenting the D1 refactor decisions and scope split (D1/D2).
doc/source/reference/linq_fold_patterns.rst	Updates pattern documentation to reflect new group_by pattern rows and reducer dispatch mechanism.
benchmarks/sql/results.md	Refreshes benchmark matrix rows for group_by/join_groupby and updates the generation metadata.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated no new comments.