[AdaWorldAPI]

What

hubspo-store::log::Log::history(specs, row, col) (feature lance) returns every write of one property on one record, oldest first:

• the value given at insert (absent included);
• each update;
• the record's deletion.

Each entry carries its version and the owner the batch wrote on behalf of (Entry { version, owner, change: Change::{Set(Option<Value>), Deleted} }).

There is no history store; history is the log. Each log batch runs one Quack program, row = r ∧ ((lane = c ∧ op = SET) ∨ op = DELETE), on mask-risc over the log's Arrow buffers. The buffers are borrowed, not copied, and only matching records are decoded.

To make that one program over three U32 lanes, the log's op and lane columns widen to u32. The format shipped in #23 and no data has been written with it, so nothing needs migrating.

The log is never compacted (decision 2026-10-10). The append order is the version order, so history and replay no longer sort. History's sort was inert under a disable run. Replay still refuses an out-of-order log, as a batch whose version is not the next one. The decision is recorded in plan §5.2.

The board marks D-STO-1 Shipped and D-STO-5 In progress.

Evidence

Tests (cargo test -p hubspo-store --features lance):

• For every record and both lanes over 30 random batches, the history equals one written down independently from the batches. Anti-vacuity: the comparison covers histories with updates, deletions, negative values and absent values.
• An unknown lane is refused.
• An empty log reads empty.

Disable runs:

disabled guard	result
row = r term	red
op = DELETE arm	red
op = SET term (BORN/DELETE leak in)	red
I32 sign bit	red
sort by version	green, and now removed

The budget fails, and it is stated, not hidden

Measured with examples/probe_history.rs: release, 100k records, 1,000 versions × 1,000 updates, 1.2M log records, one fragment per version.

• History reads in 336 ms, of which the raw scan is 320 ms.
• P-HIST-1's budget is < 20 ms, so this fails it.
• The scan is the cost; the program is 3–15 ms of it.

ISS-HIST-SCAN-COST lists the levers:

1. Projecting only row, lane and op.
2. An index on row. P-SPARSE-1 measured 16.9 ms with Lance's BTree on an uncompacted log of 1,000 fragments, which is inside the budget. That index goes through Lance's internal DataFusion. Whether it fits iron rule 5 is an operator decision, and the plan's alternative is our own change index.

D-STO-5 stays open until one of these lands.

Local: fmt, clippy (-D warnings, with and without lance) and the workspace tests pass. No other HubSPO-rs PR is open.

🤖 Generated with Claude Code

https://claude.ai/code/session_01PPcakjk7ioQsnasmNDtc5z

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-10T18:10:19.074510Z	de52dd3	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: de52dd31db

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".