📐world-store-order

The groupant beside world/src/store.rs in the Mova repository: what the code computes, what it never does, how to go back from its outputs, its invariants, cost and checks.

Facts

Components

Obernants

The typed relations around the forward model: the behaviour complement (what the code never does), the reverse reconstruction (every input compatible with an output) and specialised views.

Model text

descriptant

Shuffled blocks

Forward

Scan metadata, cut every book boundary or a paragraph boundary after the size threshold. Hash sentence count, books-file length, first version and UTC day. Reuse matching cache or shuffle with xorshift Fisher-Yates, then write order.bin and order.tsv directly. Prefix emits whole blocks until its requested sentence threshold is reached.

Invariant

Fresh blocks partition metadata sentence IDs. Reused order.bin is trusted and not validated; truncated data can panic. Cache identity excludes size and corpus content beyond coarse measurements. This is a daily order, not cross-day reproducibility.

Cost

Fresh order scans n metadata entries and shuffles b blocks. Cached order reads b records. Prefix expands selected blocks and reserves the requested capacity, which can exceed n; bounded finite percentages are assumed for safe probes.

Epoch memory

2026-10-29: a 1% prefix over six sentences in two three-sentence books emits three IDs. The exact-budget false model is a negative control. A size change after cache creation does not trigger rebuilding.

catalog

world-store-order: typed model catalog

Domain: finite valid byte files, bounded IDs and successful allocation; an operation's arguments plus its initial filesystem and scheduler state are inputs. Observations include returns/errors and written bytes. No simultaneous writer, path traversal, malformed offsets or crash is assumed unless explicitly modeled. Source-derived conditional relation, not a proof of all Rust executions. Evidence: tests.md distinguishes executed probes from proposed oracles.

Every entry is a separate observation/constraint. Status: partial source-derived conditional model. Subject: world/src/store.rs at SHA-256 6801677c4f17273e615090a790f2e4018387ca5d006aab899bafda09f01b24d9. Reverse operators retain all compatible antecedents; complements are candidate inner bounds, not complements of an empirically fitted forward function. Proposed oracles are not executed evidence.

nonbehavior

Type: behavior_complement.

U is valid metadata and size x block list. U minus R contains a positive block crossing a book boundary or splitting a paragraph solely because size was reached. Candidate inner bound: book boundary always cuts; size only cuts at paragraph end. Probe explicit paragraph fixture.

antecedents

Type: reverse_reconstruction.

For observed block spans, every metadata sequence whose internal positions avoid the cut predicate and whose block ends satisfy book_end or(size reached and para_end) is a preimage. Book and paragraph numeric labels are not uniquely recovered. Oracle: relabeling preserves blocks.

partition

Type: conservation.

For n>0, blocks partitions [0,n) into contiguous positive intervals; total length n. Fisher-Yates permutes these intervals, so expanded valid order covers each ID once. n=0 gives no blocks. Probe sizes 0,1,2,400 checks sum, contiguity and expansion.

budget

Type: resource_preimage.

Prefix wants ceil(n*pct/100) but emits whole blocks: for 0<want<=n, want<=out_len< want+last_block_len; want>n emits n. pct<=0 casts to zero; NaN also becomes zero; infinity may cause capacity failure and is outside admitted probe domain. Probe bounded percentages and overshoot.

This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/world/src/store.rs.groupant/order.groupant/