Elamur Groupantsmova-elephant-merge
The groupant beside tools/elephant/src/merge.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 | Forward approximant |
| Ast | partial | syntax.json |
| Composition | partial | Composition |
| Cost | partial | Cost |
| Descriptant | present | Descriptant |
| Invariant | partial | Invariant |
| Tests | finite evidence | 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 tamping merge: parts or runs, each added or subtracted, into one canonical run (the signed counts restricted to positive ones, in key order, with exactly the OOV words its records use). Source: tools/elephant/src/merge.rs.
Forward approximant
k-way merge of the OOV tables -> union ranks, monotone rank maps and signed frequency sums (no word held); merge pass 1 (binary heap over sorted streams, keys remapped, signed sums): negative -> error, zero -> dropped, positive -> counted and OOV words marked; canonical table = marked words, streamed again from the inputs into the output; merge pass 2 sums in union ranks and writes records remapped into canonical ranks.
Invariant
Counts form a commutative group, so compact is invariant under the order and grouping of its inputs (associativity, commutativity) and subtraction is exact ((a+b)-a = canonical b); a merged count above u32 or below 0 is refused; inputs must share the store's format.
Cost
Two streaming reads of the records, O(R log k) for R records and k inputs, and two of the OOV tables; memory 4 bytes per input OOV entry plus 9 bytes per union word, no strings or records held; one write. Measured 2026-11-06: 4 runs + 5 parts (35.1 M records out) in 21-23 s at 0.04 GB; 11 whole-Wikipedia parts (126.0 M records in, 77.8 M out, 2.0 M OOV words) in 69 s at 0.10 GB.
Composition
Parent: elephant. Called by store::tamp.
Behavior complement
Never holds the records in memory; never mixes formats; never writes a non-positive count. Evidence status: source-derived model with finite checks, not a proof.
Reverse reconstruction
Given the output and all inputs but one, the missing input's positive counts are recoverable by subtraction; the merge tree is not recoverable (it does not matter, by the invariant).
Tests
Checks
merge::tests: order-free bytes, exact subtraction, retraction of what was not added refused, a format-1 input in a format-2 merge refused; lsm: two merge trees on the 50 000-article pass write the same 143 MB run. These are finite fixtures and measurements; not a proof.
History
2026-11-06 · document
Epoch 2026-11-06
First model, written with the source (the Elephant made incremental and an animal on Mova).
Provenance
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/tools/elephant/src/merge.rs.groupant/