feat(dapi): composite document queries on the getDocuments V1 wire

[QuantumExplorer]

Replaces #4599, which GitHub closed as merged when a mistaken force-push briefly collapsed the stack's branches onto one commit. Same branch, same content, same base.

Issue being fixed or feature implemented

Second PR of the composite-query stack (core in #4598): puts composite document queries on the typed getDocuments V1 wire and serves them from drive-abci. A composite query is a page plus sub-queries derived from its proven results (joins, lookups, counts, siblings), answered as one merged proof; the motivating case is a social feed whose per-page enrichment is today a burst of dependent DAPI calls.

What was done?

Wire (platform.proto)

• GetDocumentsRequestV1.sub_queries (field 14, repeated SubQuery): each sub-query carries its own contract (empty = the page's), document type, fixed where_clauses / order_by, optional limit, a Kind (DOCUMENTS | COUNT) and an optional Binding { source, source_property, field } whose source is 0 for the page or n for sub_queries[n-1]. Presence selects composite mode; the request's own clauses describe the page. chained and sub_queries are mutually exclusive. No CBOR anywhere: typed clauses only, like the rest of V1.
• ResultData.composite (variant 7, CompositeDocuments): the page documents plus one SubQueryResult per sub-query (Documents or CountEntries) on the no-proof path. The proof path is the single merged proof in the standard envelope.
• build.rs: serde default on sub_queries so mock vectors captured before the field stay wire-compatible (same rule as chained).

drive-abci (dispatch/composite.rs)

Routes before the SELECT machinery. Gates: page limit required in [1, max_query_limit]; selects empty or a single DOCUMENTS projection; group_by, having, cursors, offset and time-range clauses (page or sub-query) refused; each distinct sub-query contract fetched once; kinds, limits and bindings decoded into drive's typed shapes; drive's validate failures surface as query errors. Proof path: query_composite_documents_with_proof under the standard proof envelope. No-proof path: documents serialized with their own type and contract, counts as CountEntry rows.

An old node ignores field 14 (proto3 unknown field) and serves the plain page; the verifier in #4598 refuses that page-only proof whenever a sub-query derived anything.

Mechanical: every existing GetDocumentsRequestV1 literal gains sub_queries: Vec::new(); the proof verifier's result-variant name table gains the composite arm.

How Has This Been Tested?

dispatch/composite.rs tests against the yappr-feed fixture plus the dashpay contract:

• should_return_the_page_and_every_sub_result_without_proof: like counts, quoted-post join, cross-contract profile lookup (a missing profile is a proven absence)
• should_prove_end_to_end_through_the_v1_wire: the wire proof verifies through DriveCompositeDocumentQuery::verify_composite_documents_proof rebuilt client-side
• should_require_an_explicit_page_limit, should_reject_chained_and_composite_together, should_reject_an_unknown_sub_query_contract, should_surface_shape_rejections_as_query_errors

The whole document_query::v1 module (103 tests, chained included) passes; clippy --all-targets -D warnings clean on drive-abci, dapi-grpc, drive-proof-verifier and dash-platform-queries; dash-sdk and wasm-sdk compile against the new proto.

Regenerated JS gRPC clients follow in a separate commit on this branch (Docker regen).

Breaking Changes

None. Additive wire fields; old clients and old nodes are unaffected (an old node answering a composite request fails closed on the client).

Checklist

• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have added or updated relevant unit/integration/functional/e2e tests
• I have added "!" to the title and described breaking changes in the corresponding section if my code contains any
• I have made corresponding changes to the documentation if needed

🤖 Generated with Claude Code

[thepastaclaw]

🕓 Ready for review — 55 ahead in queue (commit 3fed9ed)
Queue position: 56/58 · 2 reviews active
ETA: start ~2026-09-07 14:43 UTC · complete ~2026-09-07 15:38 UTC (median 54m across 30 recent reviews; 2 slots)
Queued 1d 4h ago · Last checked: 2026-09-06 14:00 UTC

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Team

Run ID: 0ac222f6-1201-4aa2-bddd-bea498a8af90

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}