gparse-eisner-dynamic-program
The groupant beside gparse/src/eisner.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 |
| Cost | symbolic source derived | Invariant |
| Descriptant | present | Descriptant |
| Invariant | partial | Invariant |
| Tests | partial 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 |
| Chart base | chart base | partial source derived model | Chart base |
| Incomplete join | incomplete join | partial source derived model | Incomplete join |
| Root constraint | root constraint | partial source derived model | Root constraint |
| Tie order | tie order | partial source derived model | Tie order |
Model text
Descriptant
Source-bound description
Span filling uses rounded f32 max-plus, stores earliest split for strict improvements, and selects one root. Exact-real maximum-tree interpretation requires exact arithmetic or candidate-order preservation.
Approximant
Forward approximant
Span filling uses rounded f32 max-plus, stores earliest split for strict improvements, and selects one root. Exact-real maximum-tree interpretation requires exact arithmetic or candidate-order preservation.
Domain: valid source-specific dimensions and resources stated in invariant; no automatic production substitution. Numerical observations are finite checks, not an end-to-end relative bound. Executable Rust surrogate: unsupported gap; existing exact tests are not surrogates. Future metric-specific caps may be 1/2/5/10%, never above 10%.
Invariant
Invariant and cost
Square scores with finite intermediates are needed for meaningful maximizing backpointers; f32 rounding can differ from real max-plus ordering.
Cost
O(n³) arithmetic, O(n²) charts/backpointers.
Allocation, finite arithmetic, platform dimensions and external failures are not universally certified.
Models
Typed model catalog
Behavior complement
A later equal candidate cannot replace earlier maximizing split/root.
Reverse reconstruction
A chosen split is compatible with candidate vectors whose rounded f32 comparisons select it; real-score inequality cones are valid only with order preservation. All -infinity candidates preserve fallback indices, which need not be first ascending.
Chart base
Complete singleton spans initialized0, others -infinity.
Incomplete join
Best C_left(s,r)+C_right(r+1,t) receives edge in either direction.
Root constraint
Only final root choice adds sc[0][r], ensuring one root child on admitted domain.
Tie order
For representable finite candidates, strict > and ascending r retain first maximum. If all candidates are -infinity, default a1=t can survive instead.
Tests
Checks and limits
Selected parent checks relevant to this fragment (not exhaustive): eisner::tests::best_tree_with_one_root; eisner::wave30::empty_singleton_and_first_root_tie; decoder_accepts_no_words
Selected commands run twice debug and twice release; exact receipts and source-to-test mapping retained in wave30 gparse external run directory. Tests do not certify allocation failure, all NaN/overflow inputs, trained-model accuracy or corpus timing. Compiled mutants cover selected boundaries; per-source coverage is recorded in the external map, not presumed.
History
2026-10-31 · descriptant
Epoch 2026-10-31
Prospective model was recorded before edits. Added tests use bounded synthetic fixtures; production algorithms unchanged. Coverage is partial, not proof. No trained corpus, live teacher or throughput experiment was executed.
Provenance
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/gparse/src/eisner.rs.groupant/dynamic-program.groupant/