Elamur Groupantstreestore-store
The groupant beside tools/treestore/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
Implementation material. A Mova code model describes this code at one source revision: partial models, finite checks and their stated scope. It is not a reviewed general exposition, and its labels are shown as written, never raised.
Components
| component | label | where |
|---|---|---|
| Approximant | partial | Approximant |
| Ast | partial | syntax.json |
| Composition | partial | Composition |
| Cost | symbolic source derived | Invariant |
| Descriptant | present | Descriptant |
| Invariant | partial | Invariant |
| Tests | executed bounded checks | Tests |
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.
| relation | type | label | where |
|---|---|---|---|
| Behavior complement | behavior complement | partial source derived model | Behavior complement |
| Reverse reconstruction | reverse reconstruction | partial source derived model | Reverse reconstruction |
Model text
Descriptant
The .mts file: magic and model id, groups of consecutive articles up to a raw-size target, each stream of a group one
frame (stored, zstd, zstd with the stream's dictionary, or the context coder β whichever is smallest: the Grand
Spider's dictionary levels at stream granularity), and a zstd footer with the groups and per article its info line,
JSON length and SHA-256 prefix. A streaming Writer (add/finish) and the batch writer give identical bytes. Epoch
2026-11-06: on tier 1 about 2.9Γ smaller than the current per-article zstd 9 records and 2.1Γ smaller than one zstd 19
--long=27 frame per shard (which has no random access).
Approximant
size(file) β Ξ£_groups Ξ£_streams min(zstd_dict, zstd, cm, raw) + 16Β·articles + footer. Read of one article = one seek, decompress its group's 16 frames, slice, decode, hash. Measured 2026-11-06 (shard 00050, 2 MB groups): 11.85 MB; random article (avg 530 KB of JSON) 25β60 ms; full decode 26β32 MB/s of JSON; packing β 4 s per 500-article shard on 6 threads, peak β 0.7 GB.
Invariant
For every written article: decode gives exactly the record (length and SHA-256 prefix checked on every read); every
frame is decompressed and compared at write time; same records, model and settings give the same file (groups are
coded in parallel but appended in order). The file appears only complete (written as .part, synced, renamed).
Composition
encode_group β codec::encode per record, then compress per stream; Shard::group β decompress; Shard::decode β codec::decode + checks. dir.rs indexes files by title.
Behavior complement
No in-place update or deletion of articles (a changed article means rewriting its file or appending a new file), no sentence-level index inside a group, no encryption, no cross-file deduplication.
Reverse reconstruction
From a file and its model every record is recovered exactly; the footer alone recovers titles, lengths and hashes but not content.
Tests
tests/roundtrip.rs: batch vs streaming writer identical, round trip of all records, negative controls (damaged bytes
in a group never return wrong bytes and are caught; a wrong model id is refused). Real-data gate: treestore verify.
Provenance
This page in Markdown β https://md.groupants.elamur.ai/free/en/mova/tools/treestore/src/store.rs.groupant/