Decide: are calculation results authoritative records or cacheable derivations? #15

New issue

Closed

opened 2026-07-17 01:17:58 +00:00 by claude-bot · 1 comment

claude-bot commented

2026-07-17 01:17:58 +00:00

Contributor

Open Question carried over from docs/roadmap.md.

Should calculation results be treated as authoritative records, or always as cacheable derivations from home data?

Consequential: if results are cacheable derivations, they can be dropped and recomputed freely, and cleanup is trivial. If they are authoritative records, they must survive formula changes, which means historical results cannot be regenerated and must be retained as-is.

The current implementation leans ''cacheable'' - results carry a calculation version and input hash and are invalidated when either changes. But report snapshots (Milestone 17) imply some results are meant to be durable records of a point in time.

Resolving this determines the retention policy question too.

Open Question carried over from `docs/roadmap.md`. > Should calculation results be treated as authoritative records, or always as cacheable derivations from home data? Consequential: if results are cacheable derivations, they can be dropped and recomputed freely, and cleanup is trivial. If they are authoritative records, they must survive formula changes, which means historical results cannot be regenerated and must be retained as-is. The current implementation leans ''cacheable'' - results carry a calculation version and input hash and are invalidated when either changes. But report snapshots (Milestone 17) imply some results are meant to be durable records of a point in time. Resolving this determines the retention policy question too.

claude-bot added this to the Open Questions milestone

2026-07-17 01:17:58 +00:00

claude-bot added the

decision

label

2026-07-17 01:17:58 +00:00

claude-bot referenced this issue

2026-07-17 03:08:11 +00:00

Make the calculation input hash cover actual state content #38

claude-bot referenced this issue

2026-07-17 03:09:14 +00:00

Realized-savings ledger: counterfactual old-plan cost vs actuals to date #45

claude-bot commented

2026-07-17 15:21:15 +00:00

Author

Contributor

Decision (Ryan, 2026-07-17): calculation results are cacheable derivations, not authoritative records.

Consequences:

Results can be dropped and recomputed freely from home data; formula changes simply invalidate the cache. No obligation to keep historical results reproducible under old formulas.
The durable point-in-time records are report snapshots (home_report_snapshots, migration 006) — anything that needs to survive formula changes should be captured as a snapshot, not by retaining raw results.
Retention (#16) can be a small window: always keep the latest completed result (see #32), cap failed rows, prune everything else freely.
#45 (realized-savings ledger) should pin its historical figures to report snapshots or recompute-on-demand, not to retained calculation rows.

Closing as decided.

**Decision (Ryan, 2026-07-17): calculation results are cacheable derivations, not authoritative records.** Consequences: - Results can be dropped and recomputed freely from home data; formula changes simply invalidate the cache. No obligation to keep historical results reproducible under old formulas. - The durable point-in-time records are **report snapshots** (`home_report_snapshots`, migration 006) — anything that needs to survive formula changes should be captured as a snapshot, not by retaining raw results. - Retention (#16) can be a small window: always keep the latest completed result (see #32), cap failed rows, prune everything else freely. - #45 (realized-savings ledger) should pin its historical figures to report snapshots or recompute-on-demand, not to retained calculation rows. Closing as decided.

claude-bot referenced this issue

2026-07-17 15:21:18 +00:00

Decide: retention policy for old calculation runs #16

claude-bot closed this issue

2026-07-17 15:21:19 +00:00