Files
aevolution/schemas/ecology/regional-cellular-census-v1.md
dingo 274f62cb89
Foundation / checks (deploy) (push) Successful in 5s
Foundation / checks (ecs) (push) Successful in 2m29s
Foundation / checks (network) (push) Successful in 20m54s
Foundation / checks (client) (push) Successful in 22m0s
Foundation / checks (tsan) (push) Successful in 27m5s
Foundation / checks (postgres) (push) Successful in 1h4m20s
Foundation / postgres-tools (push) Canceled after 0s
Foundation / server-image (push) Canceled after 0s
Foundation / image (push) Canceled after 0s
Foundation / checks (checks) (push) Canceled after 1h12m23s
feat(lifecycle): qualify census of funded maturing owners
2026-10-19 07:56:58 -04:00

5.5 KiB

Regional cellular census projection v1

Follow-on scoped census sets compare complete immutable genome records and permanent resident IDs retained privately by these observations. They require exact declared coverage and boundaries before querying multiple regions; no public row, biological codec or rule version changes.

Authority: ADR-0150 and ADR-0159, under ADR-0133. Portable API: evo::server::CellularCensus. This is an in-memory observation contract, not a persistent codec, network message or complete world evidence set.

Boundary and rows

The boundary records exact world, region, frame, completed simulation tick, 50,000 µs quantum and owner/rules/population versions. Raw observation has population format zero; exact populations retain envelope 1 or 2. Rules 8 with owner 6/7 retain their support. Rules 9 requires owner 7 and either raw or population envelope 2. Unsupported sources fail. The projection version is 1, separately from every biological/checkpoint format.

Maturation reads the exact retained body and reserve of each life. It does not substitute adult target size, count a growing organism twice or include spent construction heat. No-tick rule upgrades remain distinct committed history cuts. Existing history/snapshot readers replay each supported cut before projection; they never reinterpret a retained rules-8 cut using the current head's rules. Regional sets still reject mixed rules 8/9, even at equal ticks. Matching raw and cohort observations can coexist when every biological boundary agrees.

Rows are unique and ascending by immutable genome ID. Each retains its genome schema, development version and exact parent IDs, including ancestors with no resident carriers. These links describe genome ancestry; they do not fabricate organism parentage or a species classifier. Mutation counts use a resident's current genome rather than its immutable birth genome.

Each row contains five disjoint resident categories:

Category Meaning
individual wildlife Zero-account living resident without a cohort tag
cohort wildlife The same owned resident store, classified by an exact membership tag
active controlled Controlled living resident with active presence
lingering controlled Frozen controlled life still lingering in its world slot
suspended controlled Frozen controlled living stock outside active occupancy

Each category reports a whole-individual count, body inventory and reserve inventory, in pJ and the four disjoint C/N/P/other fg channels. There are no fractional organisms, statistical representatives or duplicate cohort inventory. Counts and inventories exclude corpses, detritus, habitat food, nursery, heat and cumulative boundary inflow/outflow. Those compartments still belong in a separate whole-region conservation audit; they are not living or viable reproductive stock.

retained_nonresident_records counts historical departed organisms by their locally recorded genome. It carries no inventory or current presence/absence meaning. A remote mutation or death need not be known by this historical source. It is never added to resident stock. Returned residents cease being nonresident under the existing migration contract.

Queries and authority limits

query(anchor, exact_genome) selects only the named genome's current residents. query(anchor, descendants) includes that anchor and all locally retained genome descendants. Child adjacency and a visited set prevent repeated counting; iterative traversal avoids native stack growth. Unknown anchors and invalid scope are errors, not empty observations. Channel/count overflow refuses a result.

The archive caps observation at 4,096 genomes and 65,536 historical organism records; the owner caps resident lives at 256. Enumeration uses retained records, never the numeric distance between sparse IDs. Building adjacency costs O(G log G + E log G); querying descendant closure costs O(G + E), with E bounded by the supported genome-parent limits. No database, clocks, threads or renderer are involved. Returned rows and boundary remain detached from subsequent simulation.

There is deliberately no terminal-extinction status. Zero carriers describes only this exact regional observation; a missing row is not a zero for another region. The API supplies no commit, owner epoch, history branch, region-set coverage, pending-transfer status or proof of current authority. A detached candidate or destination ghost can be observed for comparison but cannot be published as owned world stock. Before paired migration commit, the frozen committed source owns migrants; afterwards only the committed destination owns them.

The committed history reader derives this projection from the existing atomic biological journal after complete replay, binding its exact manifest/digest without a redundant census write. Registered snapshots additionally capture database history domain, required region coverage and pending handoffs. M12.7 must still reconcile every required viable store at a complete watermark; that global decision is outside this local projection.

ADR-0151 now binds projections on read to the existing atomic immutable biological journal through committed regional history. The commit itself is the canonical event; no redundant census write is added. Complete world/branch reconciliation remains separate.