{
  "api_version": "v1",
  "dataset_version": "AdityaNet_v2_dataset_r1",
  "dataset_sha256": "43fd0e228b28ae6bc7e468c3acf68722768bd62b73798eb6631e9e6233b71ed9",
  "generated_at": "2026-07-29T13:34:16.829920+00:00",
  "source_commit": "99af630",
  "data": {
    "title": "Reproducibility Report \u2014 `AdityaNet_v2_dataset_r1`",
    "date": "2026-07-18",
    "lead": "**Result: PASS. Rebuilding from the raw ISSDC archive reproduces the canonical tables BYTE-IDENTICALLY \u2014 12/12 sampled days, both T1 and T2, byte match AND content match.** Evidence: `reproducibility_check.json`.\n\n---",
    "sections": [
      {
        "heading": "1. What was tested",
        "body": "Twelve days were selected by even spacing across the full archive span (2024-02-01 \u2192 2026-06-15) \u2014 not cherry-picked \u2014 and rebuilt **from the raw L1 FITS**, through the frozen parsers and builders, then compared against the frozen dataset.\n\nTwo distinct claims were measured separately, because they are not the same thing:\n\n- **Content identity** \u2014 the *values* are identical. This is the scientific claim: a rebuild must reproduce the same measurements.\n- **Byte identity** \u2014 the *parquet file bytes* are identical. Stronger, but partly a property of the writer environment (parquet embeds library metadata), so it was measured rather than assumed."
      },
      {
        "heading": "2. Result",
        "body": "| Day | T1 content | T1 byte | T2 content | Rebuild |\n|---|---|---|---|---|\n| 2024-02-01 | \u2705 | \u2705 | \u2705 | 0.6 s |\n| 2024-03-10 | \u2705 | \u2705 | \u2705 | 0.4 s |\n| 2024-04-20 | \u2705 | \u2705 | \u2705 | 0.4 s |\n| 2024-05-31 | \u2705 | \u2705 | \u2705 | 0.4 s |\n| 2025-08-09 | \u2705 | \u2705 | \u2705 | 0.4 s |\n| 2025-09-30 | \u2705 | \u2705 | \u2705 | 0.4 s |\n| 2025-11-12 | \u2705 | \u2705 | \u2705 | 0.4 s |\n| 2025-12-23 | \u2705 | \u2705 | \u2705 | 0.4 s |\n| 2026-02-01 | \u2705 | \u2705 | \u2705 | 0.4 s |\n| 2026-03-15 | \u2705 | \u2705 | \u2705 | 0.4 s |\n| 2026-05-04 | \u2705 | \u2705 | \u2705 | 0.4 s |\n| 2026-06-15 | \u2705 | \u2705 | \u2705 | 0.4 s |\n\n**Content identity: T1 12/12, T2 12/12. Byte identity: T1 12/12, T2 12/12.**\n\nBoth the 340-channel spectral arrays (T2) and the scalar light-curve table (T1) reproduce exactly. Row counts match (1,440 rows/day) on every sample."
      },
      {
        "heading": "3. Determinism: why it holds",
        "body": "The pipeline contains **no non-deterministic operation** by construction:\n\n| Property | Guarantee |\n|---|---|\n| No randomness | No RNG anywhere in parsers or builders (no sampling, no shuffling, no initialisation) |\n| No imputation | NaN is preserved as the missing-data sentinel; never filled, never dropped (verified by AST scan + tests) |\n| No reordering | The parser preserves archive order losslessly; HK sorting is an explicit out-of-parser utility |\n| Order-independent resolution | The Version Resolution Engine builds a coverage map in two phases (collect \u2192 resolve); shuffled candidate order yields identical maps, proven on real data |\n| Order-independent aggregation | All minute statistics are sum/mean/max/min over finite values |\n| Fixed hashing | Provenance SHAs come from the frozen Phase 0.5.1 manifest, not recomputed at build time |\n\n**Version-resolution stability across three independent full builds:** 1,065,572 owned (minute, detector) pairs, 48,604 conflicts, R1 47,328 / R2 1,276 / R3 0 / F-14 0 \u2014 byte-stable every time."
      },
      {
        "heading": "4. The one documented exception",
        "body": "**T7 `provenance_manifest` contains `parsed_at_utc`**, a build wall-clock timestamp. This column differs between builds *by design* \u2014 it records when a build ran. Every other T7 column is deterministic.\n\nConsequently:\n- The reproducible **content hash** of T7 **excludes** `parsed_at_utc` (`NON_REPRODUCIBLE_COLUMNS` in `freeze_dataset.py`).\n- The **file SHA-256** of T7 in the manifest pins *this* frozen artifact exactly, as it does for every other file.\n\nThis is stated rather than engineered away: removing the timestamp would have modified the dataset after freezing, and the timestamp is legitimate provenance."
      },
      {
        "heading": "5. Environment",
        "body": "| Field | Value |\n|---|---|\n| Python | 3.12.12 |\n| numpy / pandas | 1.26.4 / 2.2.2 |\n| pyarrow / astropy | 16.1.0 / 6.1.7 |\n| Platform | macOS-26.5.2-arm64 (Apple Silicon) |\n| Lockfile SHA-256 | `6899e001b1c4d64a5aec01b3ee1d2cfce2a3ff38ab64f827a75ab9d55fd13d3f` |\n\n**Scope of the byte-identity claim:** demonstrated in this pinned environment. Byte identity depends on the parquet writer (`pyarrow==16.1.0`); a different pyarrow may produce different bytes for identical data. **Content identity is the portable guarantee**; byte identity is the observed, stronger result here. Both are reported so the distinction is not lost."
      },
      {
        "heading": "6. Timing",
        "body": "| Stage | Time |\n|---|---|\n| Per-day rebuild (T1 + T2, from raw FITS) | **0.4 s** median |\n| Full archive rebuild (424 SoLEXS days + 389 HEL1OS orbits) | **93.75 min** |\n| Archive extraction (827 ZIPs \u2192 135 GB) | 6.8 min |\n| Scientific validation | ~20 min |\n| Sample verification (this report) | ~6 s |"
      },
      {
        "heading": "7. How to reproduce",
        "body": "See `DATASET_MANIFEST.md` \u00a75. In short: install from the lockfile, run `extract_archive.py`, run `build_canonical.py`, then `freeze_dataset.py` and compare `dataset_hash` against `43fd0e228b28ae6bc7e468c3acf68722768bd62b73798eb6631e9e6233b71ed9`.\n\n**Verdict: the dataset is reproducible from the raw archive.** Content-identical by construction and byte-identical in the pinned environment, with one documented, deliberate exception (`parsed_at_utc`)."
      }
    ],
    "source": "artifacts/v2/phase05/REPRODUCIBILITY_REPORT.md"
  }
}