Module shape 

aver shape — module-shape presentation layer.

Stage 1 of issue #232 (0.23 “Shape”) split this file into two tiers:

• Recognition (aver::analysis::shape) — per-fn facts walk, archetype classifier, call-graph SCC. Stable typed primitives.
• Presentation (this module) — ModuleShape vector + Kind + verify report + histogram + Layer fingerprint distance + corpus walker + renderer + aver.toml [[shape.layer]] / [[shape.expected]] bridges. Everything the CLI / LSP / JSON consumers see.

The recognition primitives are re-exported below so existing callers (and the research test) keep working through this module during the transition; downstream stages will migrate them to import directly from aver::analysis::shape.

Structs

CorpusSummary

Per-Kind / per-Layer aggregate over a corpus.

FnShape

Histogram

LayerFingerprint

LayerVerdict

ModuleShape

RenderOptions

ShapeReport

VerifyReport

Enums

ApiShape

Bucket

Five-bucket histogram a Layer fingerprint is defined against. Mapped from the 14 archetype labels via archetype_to_bucket.

CorpusEntry

Outcome of analyzing a single file inside a corpus walk. Files that fail typecheck or parse get the error attached instead of being silently skipped — the corpus view should show what’s broken.

Entry

Kind

Layer

Purity

StateShape

TypeSurface

VerifyLevel

Functions

analyze_dir

Walk a directory recursively and analyze every .av file. root is also used as the default module_root for dep resolution — callers that want a different module root can pass it explicitly.

analyze_path

analyze_path_with

Same as analyze_path but lets the caller supply a custom layer fingerprint table + basis label (used by the CLI when aver.toml has [[shape.layer]] overrides).

analyze_source_with

Analyze a source string directly, without touching the filesystem for the entry file. Used by the LSP (editor buffer may differ from on-disk content) and any future embedder that holds the program in memory. module_root resolves depends [...]; file_label is the path string surfaced in the report’s file field; fallback_module_name is used when the source has no module declaration.

archetype_to_bucket

builtin_v0_layer_fingerprints

classify_layer

Nearest fingerprint by Euclidean distance over bucket %. Confidence is (max_dist - chosen_dist) / max_dist, penalized by small total_fns (< 5 fns flattens confidence regardless of fit).

derive_kind

fingerprints_from_config

Built-in v0 fingerprints sourced from the casual single-author corpus measurement that landed on the tweet. Concrete numbers:

render_corpus_json

render_corpus_text

render_json

render_text

summarize_corpus