glove-parity: multi-hop questions over the software graph (D-SAG-1-4)

[AdaWorldAPI]

Stacked on #1519; merge that first. Until it merges, this diff also shows #1519's commits.

What

SoftwareGraph::hops(start, path) follows a path of (relation, direction) hops through the software-awareness graph. Each hop returns a set: each node once, in node order.

This is the operator's reading of chained hops: the masks of one query fold into one materialisation, and each hop is filtered at the source.

• Each hop is one Quack program over the edge lanes already in memory: rel = R AND from IN (S), keeping its rows.
• The previous hop's results arrive as a value set (Filter::in_u32), sized by those results and read from the other end of the kept rows. No hop reads a plane another program produced, so the chained-mask shape the repo forbids (Quack Filter::Semijoin's precondition) is never used.
• An empty set ends the walk without running a program.
• An unknown start node is an error, never an empty answer.

This unblocks the multi-hop questions D-SAG-1 had to leave open: who else reads a field this method reads, which claims stand behind its fields, and which programs share primitives.

Evidence

tests/awareness.rs checks four paths of up to four hops against a row loop over the edges:

• co-readers;
• field → claim;
• a four-hop claim co-dependency;
• a four-hop shared-primitive path.

A second test checks that a hop path follows a reprojection, which depends on #1519's fix for defect A.

Each of these four disable runs turns the tests red:

1. the relation filter dropped;
2. the direction ignored;
3. the per-hop deduplication dropped;
4. an unknown start silently ignored.

Gates: cargo test and cargo clippy --all-targets -D warnings pass for the crate; cargo fmt is clean.

Still open

A value set still reads every edge row. Skipping straight to the rows of the previous hop's results needs either:

• a Lance B-tree filter() (D-TELEPORT-2, arm Db), or
• row ranges over a projection sorted by key, which waits on ISS-WITNESSED-RANGE-DOES-NOT-ATTEST-PLANE-ORDER.

Also in this push: the entries index row for #1519's D-SAG-1R entry is regenerated. --write keeps an existing row's id, so the retitled entry still read as D-SAG-1. The same fix is on #1519.

🤖 Generated with Claude Code

https://claude.ai/code/session_011BMJnDBAfzcuQh5HeT39Rh

---

Generated by Claude Code

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Warning

Billing warning: we have not been able to collect payment for this subscription for more than 72 hours. Please update the payment method or pay any pending invoices in Billing to avoid service interruption.

---

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