aggregate: ingest proven-tests JSON run reports (new input kind) (#152)

## Context

`proven-tests-and-benches` emits a JSON run report (schema_version 1)
grading each test by three-tier provenance. This adds a self-contained
`aggregate` input kind that parses that report and folds it in using the
**honest** tier→verdict mapping fixed in that repo's
`docs/INTEROP-PANIC-ATTACK.adoc`.

## Change

New module `src/aggregate/proven_tests.rs` (registered from
`src/aggregate/mod.rs`):

- Serde structs for the schema-v1 report (`ProvenTestsReport`, `Entry`,
`Summary`, `ProofStep`).
- `ReportTier::to_verdict()` — the honest mapping:
- **Actually-Proven → `Closed`** (carries a proof ladder;
`actually_proven_ladder_files()` extracts the distinct cited `.idr`
paths, which are the natural `aggregate --proof` inputs).
- **Provisionally-Proven / Unproven → `None`** — no proof artifact,
deliberately **not** read as `Holes`. Absence of a proof object is not
the presence of a hole.
- Rejects unknown `schema_version`; tolerates unknown tiers (surfaces
them in `notes()` rather than panicking).

The trust boundary is unchanged: verdicts are conditioned on the Idris
checker's soundness; panic-attack does not re-check proofs. No CLI
wiring in this change — the parser + mapper is the substantive, testable
core; the ordinary `aggregate --proof <ladder-file>` path already
handles the export direction (verified end-to-end from the proven-tests
side).

## Verification

```
cargo test --lib aggregate::proven_tests   # 5 passed
```
Tests cover: schema-v1 parse, honest tier mapping, distinct-ladder
extraction, unknown-schema rejection, unknown-tier tolerance.

Related: `hyperpolymath/proven-tests-and-benches#13`.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

https://claude.ai/code/session_01SW76f8XBgwsiGXU9x5vmQU

---
_Generated by [Claude
Code](https://claude.ai/code/session_01SW76f8XBgwsiGXU9x5vmQU)_

Co-authored-by: Claude <noreply@anthropic.com>

Author: hyperpolymath

src/aggregate/mod.rs

@@ -50,6 +50,9 @@ use std::fs::{self, File};
 use std::io::Read;
 use std::path::{Path, PathBuf};
 
+/// Ingest a `proven-tests` JSON run report as an aggregate input kind.
+pub mod proven_tests;
+
 /// Upper bound on the text we scan for verdict / coverage classification.
 /// The *hash* always covers the whole file (see [`hash_file`]); this cap
 /// only bounds the in-memory text scan for markers.

src/aggregate/proven_tests.rs

@@ -0,0 +1,245 @@
+// SPDX-License-Identifier: MPL-2.0
+// SPDX-FileCopyrightText: 2026 Jonathan D.A. Jewell <j.d.a.jewell@open.ac.uk>
+
+//! Ingest a `proven-tests` JSON run report (schema_version 1) as an aggregate
+//! input kind.
+//!
+//! `proven-tests-and-benches` grades each test by three-tier provenance
+//! (Actually-Proven / Provisionally-Proven / Unproven), where Actually-Proven
+//! carries a proof ladder citing machine-checked theorems in real `.idr` files.
+//! This module folds such a report into aggregate's vocabulary using the honest
+//! mapping fixed in that repo's `docs/INTEROP-PANIC-ATTACK.adoc`:
+//!
+//!   * **Actually-Proven** — has a proof artifact (the ladder files). Expected
+//!     to reconcile as [`ProofVerdict::Closed`]. Its cited `proof_file`s are the
+//!     natural inputs to the ordinary `aggregate --proof` path.
+//!   * **Provisionally-Proven** — type-safe framework + example/property
+//!     evidence, but *no proof object*. Deliberately NOT treated as
+//!     [`ProofVerdict::Holes`]: absence of a proof artifact is not the presence
+//!     of a hole. Recorded as informational only.
+//!   * **Unproven** — no artifact; informational only.
+//!
+//! The trust boundary is unchanged: panic-attack does not re-check the proofs;
+//! a `Closed` reading is conditioned on the Idris checker's soundness.
+
+use super::ProofVerdict;
+use anyhow::{Context, Result};
+use serde::{Deserialize, Serialize};
+use std::collections::BTreeSet;
+use std::path::Path;
+
+/// A cited proof step (only Actually-Proven entries carry these).
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct ProofStep {
+    pub description: String,
+    #[serde(default)]
+    pub proof_file: Option<String>,
+    #[serde(default)]
+    pub theorem: Option<String>,
+}
+
+/// One graded test entry in the run report.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct Entry {
+    pub test_id: String,
+    #[serde(default)]
+    pub category: Option<String>,
+    #[serde(default)]
+    pub aspect: Option<String>,
+    /// "Actually-Proven" / "Provisionally-Proven" / "Unproven".
+    pub provenance: String,
+    #[serde(default)]
+    pub proof_ladder: Vec<ProofStep>,
+}
+
+/// The run-report summary block.
+#[derive(Debug, Clone, Serialize, Deserialize, Default)]
+pub struct Summary {
+    #[serde(default)]
+    pub passed: usize,
+    #[serde(default)]
+    pub total: usize,
+    #[serde(default)]
+    pub actually_proven: usize,
+    #[serde(default)]
+    pub provisionally_proven: usize,
+    #[serde(default)]
+    pub unproven: usize,
+    #[serde(default)]
+    pub covered_cells: usize,
+    #[serde(default)]
+    pub cat_aspect_cells: usize,
+    #[serde(default)]
+    pub lattice_covered: usize,
+}
+
+/// A parsed `proven-tests` run report.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct ProvenTestsReport {
+    pub schema_version: u32,
+    #[serde(default)]
+    pub suite: String,
+    #[serde(default)]
+    pub summary: Summary,
+    #[serde(default)]
+    pub entries: Vec<Entry>,
+}
+
+/// The three provenance tiers, parsed from the report's string labels.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum ReportTier {
+    ActuallyProven,
+    ProvisionallyProven,
+    Unproven,
+}
+
+impl ReportTier {
+    /// Parse the report's tier label (the exact `Show ProvenStatus` strings).
+    pub fn parse(s: &str) -> Option<Self> {
+        match s {
+            "Actually-Proven" => Some(ReportTier::ActuallyProven),
+            "Provisionally-Proven" => Some(ReportTier::ProvisionallyProven),
+            "Unproven" => Some(ReportTier::Unproven),
+            _ => None,
+        }
+    }
+
+    /// The honest aggregate reading of a tier. Only Actually-Proven maps to a
+    /// proof verdict (`Closed`); the weaker tiers carry no artifact, so they
+    /// map to `None` rather than being misread as `Holes`.
+    pub fn to_verdict(self) -> Option<ProofVerdict> {
+        match self {
+            ReportTier::ActuallyProven => Some(ProofVerdict::Closed),
+            ReportTier::ProvisionallyProven | ReportTier::Unproven => None,
+        }
+    }
+}
+
+impl ProvenTestsReport {
+    /// Parse a run report from JSON text.
+    pub fn from_str(text: &str) -> Result<Self> {
+        let report: ProvenTestsReport =
+            serde_json::from_str(text).context("parsing proven-tests run report")?;
+        if report.schema_version != 1 {
+            anyhow::bail!(
+                "unsupported proven-tests report schema_version {} (expected 1)",
+                report.schema_version
+            );
+        }
+        Ok(report)
+    }
+
+    /// Parse a run report from a file path.
+    pub fn from_path(path: &Path) -> Result<Self> {
+        let text = std::fs::read_to_string(path)
+            .with_context(|| format!("reading {}", path.display()))?;
+        Self::from_str(&text)
+    }
+
+    /// The distinct `.idr` proof-ladder files cited by Actually-Proven entries.
+    /// These are the natural inputs to the ordinary `aggregate --proof` path.
+    pub fn actually_proven_ladder_files(&self) -> Vec<String> {
+        let mut files: BTreeSet<String> = BTreeSet::new();
+        for entry in &self.entries {
+            if ReportTier::parse(&entry.provenance) == Some(ReportTier::ActuallyProven) {
+                for step in &entry.proof_ladder {
+                    if let Some(f) = &step.proof_file {
+                        files.insert(f.clone());
+                    }
+                }
+            }
+        }
+        files.into_iter().collect()
+    }
+
+    /// A one-line, honest note per entry, applying the tier→verdict mapping.
+    pub fn notes(&self) -> Vec<String> {
+        self.entries
+            .iter()
+            .map(|e| match ReportTier::parse(&e.provenance) {
+                Some(ReportTier::ActuallyProven) => format!(
+                    "{}: Actually-Proven → Closed ({} ladder step(s))",
+                    e.test_id,
+                    e.proof_ladder.len()
+                ),
+                Some(ReportTier::ProvisionallyProven) => {
+                    format!("{}: Provisionally-Proven → informational (no proof artifact)", e.test_id)
+                }
+                Some(ReportTier::Unproven) => {
+                    format!("{}: Unproven → informational (no proof artifact)", e.test_id)
+                }
+                None => format!("{}: unrecognised provenance tier {:?}", e.test_id, e.provenance),
+            })
+            .collect()
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    const SAMPLE: &str = r#"{
+      "schema_version": 1,
+      "suite": "proven-tests",
+      "summary": {"passed": 3, "total": 3, "actually_proven": 1,
+                  "provisionally_proven": 1, "unproven": 1,
+                  "covered_cells": 35, "cat_aspect_cells": 238, "lattice_covered": 35},
+      "entries": [
+        {"test_id": "tropical-laws", "category": "Proof-Regression",
+         "aspect": "Dependability", "provenance": "Actually-Proven",
+         "proof_ladder": [
+           {"description": "comm", "proof_file": "src/ProvenTests/Tropical.idr", "theorem": "oplusComm"},
+           {"description": "assoc", "proof_file": "src/ProvenTests/Tropical.idr", "theorem": "oplusAssoc"}
+         ],
+         "result": "passed"},
+        {"test_id": "typesafe-tropical", "provenance": "Provisionally-Proven",
+         "proof_ladder": [], "result": "passed"},
+        {"test_id": "smoke-show", "provenance": "Unproven",
+         "proof_ladder": [], "result": "passed"}
+      ]
+    }"#;
+
+    #[test]
+    fn parses_schema_v1() {
+        let r = ProvenTestsReport::from_str(SAMPLE).expect("parse");
+        assert_eq!(r.schema_version, 1);
+        assert_eq!(r.entries.len(), 3);
+        assert_eq!(r.summary.actually_proven, 1);
+        assert_eq!(r.summary.covered_cells, 35);
+    }
+
+    #[test]
+    fn tier_mapping_is_honest() {
+        // Only Actually-Proven becomes a proof verdict; the weaker tiers do NOT
+        // become Holes.
+        assert_eq!(
+            ReportTier::ActuallyProven.to_verdict(),
+            Some(ProofVerdict::Closed)
+        );
+        assert_eq!(ReportTier::ProvisionallyProven.to_verdict(), None);
+        assert_eq!(ReportTier::Unproven.to_verdict(), None);
+    }
+
+    #[test]
+    fn extracts_distinct_actually_proven_ladder_files() {
+        let r = ProvenTestsReport::from_str(SAMPLE).unwrap();
+        let files = r.actually_proven_ladder_files();
+        // two ladder steps cite the same file → one distinct entry
+        assert_eq!(files, vec!["src/ProvenTests/Tropical.idr".to_string()]);
+    }
+
+    #[test]
+    fn rejects_unknown_schema_version() {
+        let bad = SAMPLE.replace("\"schema_version\": 1", "\"schema_version\": 2");
+        assert!(ProvenTestsReport::from_str(&bad).is_err());
+    }
+
+    #[test]
+    fn unknown_tier_is_flagged_not_panicked() {
+        assert_eq!(ReportTier::parse("Sorta-Proven"), None);
+        let weird = SAMPLE.replace("\"Unproven\"", "\"Sorta-Proven\"");
+        let r = ProvenTestsReport::from_str(&weird).unwrap();
+        // notes() surfaces the unrecognised tier rather than crashing
+        assert!(r.notes().iter().any(|n| n.contains("unrecognised provenance tier")));
+    }
+}