--- title: "graph-language-encoding — Mova code model" description: "Mova code model of models/graph-language-encoding.groupant: components with their labels, obernants and the model text." url: "https://groupants.elamur.ai/free/en/mova/models/graph-language-encoding.groupant/" --- Elamur: [Mova](https://md.mova.elamur.ai/free/en/) · [Dictionary](https://md.dictionary.elamur.ai/free/en/) · [Research](https://md.research.elamur.ai/free/en/) · **Groupants** This page as HTML: # graph-language-encoding The groupant beside models/graph-language-encoding.groupant 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 - **kind:** groupant_group - **status:** partial descriptive instance - **bound sources:** 8 - **repository folder:** [`models/graph-language-encoding.groupant`](https://github.com/ElamurAI/mova-free/tree/main/models/graph-language-encoding.groupant) - **repository commit:** `855da7bd4a7b` > **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 | [forward](#forward) | | **ast** | not materialized | | | **composition** | partial | [composition](#composition) | | **cost** | symbolic source derived | [cost](#cost) | | **descriptant** | present | [descriptant](#descriptant) | | **invariant** | partial | [invariant](#invariant) | | **tests** | see source local evidence | [checks](#checks) | ## 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. | id | type | label | where | |---|---|---|---| | `behavior-complement` | behavior_complement | partial source derived model | [behavior-complement](#behavior-complement) | | `reverse-reconstruction` | reverse_reconstruction | partial source derived model | [reverse-reconstruction](#reverse-reconstruction) | | `grammar-transport` | representation_transport | partial source derived model | [grammar-transport](#grammar-transport) | | `identity-boundary` | identity_transport | partial source derived model | [identity-boundary](#identity-boundary) | | `completeness-vs-realization` | observation_projection | partial source derived model | [completeness-vs-realization](#completeness-vs-realization) | | `effect-frontiers` | effect_boundary | partial source derived model | [effect-frontiers](#effect-frontiers) | | `admitted-identity-composition` | persistence_composition | partial source derived model | [admitted_identity_composition](#admitted-identity-composition) | | `identifier-history-preimages` | identity_preimage_product | partial source derived model | [identifier_history_preimages](#identifier-history-preimages) | ## Model text ### descriptant #### graph-language-encoding: wave29 Actual module declarations in kg/src/main.rs connect cols/gram/id/morph/uk/textgen; Morph and uk share Gram; textgen consumes uk::Fact and Morph synthesis; main writes Registry IDs and Edges using relation.local and full vertex Id. Registry keys are append IDs, not lexical concept indexes. global/Cargo.toml has no runtime dependencies and kg/Cargo.toml has no global dependency. Compile seed path is a separate tagged branch, not a claimed graph pipeline. Tests exercise actual Morph→parse→TSV and Fact→generate joins on invented inputs. This source-bound partial group organizes existing Mova code. Native evidence and excluded effects are recorded in source sidecars and the root wave29 report. Registration is not a universal proof or whole-system runtime claim. 2026-10-31 wave38: [scoped child-composition relations](https://md.groupants.elamur.ai/free/en/mova/models/graph-language-encoding.groupant/#wave38) distinguish admission, persistence, strict-consumer framing and history preimages. Integrated validation and its limits are recorded in the wave38 receipt; these additions do not certify every parent branch. ##### Paired composition memory — 2026-11-01 The persistence vidtvorant and decoder-fiber/admission obernants jointly explain the key and heap repairs: exact admitted identity depends on each codec's own domain, actual save/reopen and state-preserving rejection. Child key policy cannot be copied to heap labels without losing valid TAB/interior-CR names. Parent reconstruction also cannot recover overwritten byte histories or lookup history from stored IDs alone. See the [source-grounded cases](https://github.com/ElamurAI/mova-free/blob/main/docs/mathematical-modeling.md); retain the wave38 premises and historical receipts rather than promoting them to a whole-graph proof. ### model ## graph-language-encoding ### forward Tagged disjoint relation G = KG ⊎ GlobalBuild. KG: dictionary/vocabulary --Morph::load(Gram projection)--> analysis/synth maps; text+names --uk::parse--> Fact or Reject; Fact+Morph --render_check--> multiset-completeness result; Fact+Morph+Freq+OrderModel --generate--> optional sentence. In main ingestion each accepted Fact encodes one provenance and one edge/member through Registry IDs and Edges, sorted by(src,rel,dst). GlobalBuild separately maps fenced seed/optional TSV inputs to lexical concept-indexed Rust tables and expressions output. No adapter from GlobalBuild indices to kg Registry IDs, or persisted Edges back to Fact, is implemented in these sources. ### invariant For valid aligned columns n<2^32 and single-writer valid registry state, per-member lemma/relation/grammar/flags survive encoding and codec roundtrip, with stable prior IDs after save/reopen. Word coordinates and member order are not persisted. Registry checked exhaustion returns error before mutation. render_check success constrains normalized word multiset only; not semantic roles or correct inference. All claims scoped to admitted representations and finite synthetic witnesses. ### cost KG costs compose two dictionary decompressions plus per-text parsing with n<=8 but unbounded analyses, registry scans/hash lookups, E log E edge sorting and compression; optional generation O(h^2*2^h),1<=h<=12. Ingestion retains corpus/Facts/edges. GlobalBuild cost O(B+C log C+LC+R log R) plus duplicate scans and generated text; effect/resource costs are separate, not benchmarked. ### composition Actual module declarations in kg/src/main.rs connect cols/gram/id/morph/uk/textgen; Morph and uk share Gram; textgen consumes uk::Fact and Morph synthesis; main writes Registry IDs and Edges using relation.local and full vertex Id. Registry keys are append IDs, not lexical concept indexes. global/Cargo.toml has no runtime dependencies and kg/Cargo.toml has no global dependency. Compile seed path is a separate tagged branch, not a claimed graph pipeline. Tests exercise actual Morph→parse→TSV and Fact→generate joins on invented inputs. ### domain Finite UTF-8 synthetic text, unchanged tiny bz2 dictionary between passes, valid parser-derived Gram,1..8 word parse gate,1..12 explicit generation heads, aligned edge vectors below2^32, sorted src for out, valid unique registry files under single writer. Build fixtures use isolated temporary seed/out directories with absent optional closed tables. Resource exhaustion outside tested domain. ### limitations No live datasets/DB/teacher/GPU/held-out execution. Tatoeba ingestion/show/stats and nonempty optional build tables remain unexecuted. Hash-map frequency tie order in stats is not deterministic; float/u32 frequency overflow and nontransactional persistence have no new guarantee. Exact categorical semantics have no useful executable percentage approximant established. ### checks - 15 kg named outcomes twice debug/twice release, normalized outcomes equal; 3 standalone global build outcomes twice debug/twice optimized (2 substantive plus child driver). - Eight compiled mutations, compile exit0/test exit101; byte-exact restoration and complete positive rerun. - Old-code max-global-heap synthetic witness panicked overflow; corrected test passes debug/release and checks no state mutation. - Eight parent source syntax projections generated twice byte-identically;12 scoped child ranges hash-bound to parent source. - Exact native command specs and per-source mappings retained as native-commands.json/source-test-map.json. ### catalog ### behavior-complement KG rejected parser gates emit no Fact; successful render excludes mismatched normalized multiset only. Mask filtering excludes any missing-bit row; codecs exclude value loss on encoder-domain streams. Checked registry exhaustion excludes ID wrap/heap insertion. GlobalBuild success excludes duplicate canonical concept/link codes but does not exclude ignored blocks or partial write effects on failure. These exclusions are tag-specific and cannot transfer to absent KG↔GlobalBuild adapter. ### reverse-reconstruction KG output edges+Registry admit antecedent Facts with same member lemma/relation/gram but different lost word coordinates/order; filters admit arbitrary rejected rows; sorted graphs admit input permutations. Realization sentence admits multiple Fact/dictionary/count antecedents. GlobalBuild table preimage admits whitespace/comments/ignored or excluded input and absent/unreadable optional files. No unique source-text or registry reconstruction from lexical global index is implied. ### grammar-transport Morph Gram::from_vesum keys meet uk analysis grammar and textgen synth grammar after NEG/IMPLICIT masking. Implicit subject copies predicate person/number/gender; main stores fact gram in every gs and member gram in gd. This preserves packed bits, not original VESUM tags, discarded xp tags in Gram, or semantic correctness. ### identity-boundary Registry::Id=(heap<<32)|local is split exactly; relation edges use local only and require separate rel heap context. Global concept IDs are lexical vector positions with no heap/local encoding. Equating these numbers would fail even for a one-concept compiler and first global heap2^27. ### completeness-vs-realization render_check sorts normalized words; generate chooses conditional/global-frequency variants and DP order. Therefore render_check(text,Fact)=Ok does not imply generate(Fact)=text, nor does generation establish role correctness. Synthetic tests separately witness each actual function. ### effect-frontiers sub bad mask precedes reads; parse_lines loads input/names/Morph before output creation; ingestion creates names before Registry open and writes columns before registry save. Build expressions writes before global.rs. Success/failure does not compose into a transaction across these files. ### wave38 ## Registry composition models These source-derived relations refine the registry branch of this parent. Other grammar, edge and GlobalBuild branches keep their existing domains. The manifest binds the actual child revisions; no composed semantic Rust proof or transaction guarantee is inferred. ### admitted_identity_composition Type: persistence_composition. Observe the exact heap label n, key k and full identifier I across heap(kind,n) -> id(h,k) -> save -> open -> lookup. Domain: UTF-8 labels H={n: no LF and no terminal CR}, keys K={k: no TAB, LF or CR}; existing heap names/numbers and key rows are unique, well-formed and consistent, local indices fit u32, one writer and successful actual writes/flushes. Imported data outside these premises is not certified. For an admitted new or existing pair, I=(h<<32)|l, with heap(I)=h and local(I)=l. Heap encoding keeps the entire third TSV field, while key encoding keeps the remainder after the first TAB. Both codecs preserve tabs in that final field; K is deliberately stricter than codec invertibility. Their admitted encodings retain n and k after reopen, so the name/number and key/local lookups return the same I. A second save with no new rows appends nothing. This composition retains byte distinctions between canonically equivalent Unicode strings. Failure is not a multi-file rollback: name/key refusal precedes their own insertion, but an earlier successful heap creation can remain in memory if a later key call fails. save writes metadata and key stages separately. No crash, flush-failure or concurrency guarantee follows. Actual-source witnesses: id::wave38::unusual_exact_heap_names_reopen_with_stable_numbers and id::wave37::admitted_unicode_keys_reopen_with_exact_ids. Their separate success supports the scoped composition; it does not prove every graph ingestion adapter. ### identifier_history_preimages Type: identity_preimage_product. Observe persisted metadata, a full I and its decoded label/key. Under the unique valid registry domain above, require a registered h and an allocated local index 1<=l