machinechecked 0.1.15__tar.gz → 0.1.17__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (77) hide show
  1. {machinechecked-0.1.15 → machinechecked-0.1.17}/PKG-INFO +1 -1
  2. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/machinechecked.egg-info/PKG-INFO +1 -1
  3. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/machinechecked.egg-info/SOURCES.txt +3 -2
  4. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/audit.py +8 -1
  5. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/cli.py +3 -0
  6. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/config.py +8 -1
  7. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/export.py +2 -1
  8. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/manifest.py +2 -2
  9. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/paper.py +52 -3
  10. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/pipeline.py +8 -1
  11. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/published.py +42 -0
  12. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/simple.py +88 -2
  13. machinechecked-0.1.17/engine/python/prooflens/versions.py +171 -0
  14. machinechecked-0.1.15/engine/python/prooflens/web/assets/index-B_fzeuBg.js → machinechecked-0.1.17/engine/python/prooflens/web/assets/index-B1CkNBtR.js +26 -26
  15. machinechecked-0.1.15/engine/python/prooflens/web/assets/index-e4I0NErR.css → machinechecked-0.1.17/engine/python/prooflens/web/assets/index-DKOP2F1S.css +1 -1
  16. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/index.html +2 -2
  17. {machinechecked-0.1.15 → machinechecked-0.1.17}/pyproject.toml +1 -1
  18. {machinechecked-0.1.15 → machinechecked-0.1.17}/LICENSE +0 -0
  19. {machinechecked-0.1.15 → machinechecked-0.1.17}/README.md +0 -0
  20. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/machinechecked.egg-info/dependency_links.txt +0 -0
  21. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/machinechecked.egg-info/entry_points.txt +0 -0
  22. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/machinechecked.egg-info/requires.txt +0 -0
  23. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/machinechecked.egg-info/top_level.txt +0 -0
  24. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/__init__.py +0 -0
  25. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/__main__.py +0 -0
  26. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/assess.py +0 -0
  27. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/bundle.py +0 -0
  28. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/cache.py +0 -0
  29. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/data/ProofLensProbe.lean +0 -0
  30. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/data/prooflens.sty +0 -0
  31. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/data/schema/evidence.schema.json +0 -0
  32. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/data/schema/inventory.schema.json +0 -0
  33. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/data/schema/manifest-claim.schema.json +0 -0
  34. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/data/schema/manifest-index.schema.json +0 -0
  35. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/data/statuses.json +0 -0
  36. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/deps.py +0 -0
  37. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/evidence.py +0 -0
  38. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/formalization.py +0 -0
  39. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/pdfclaims.py +0 -0
  40. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/probe.py +0 -0
  41. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/promote.py +0 -0
  42. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/repin.py +0 -0
  43. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/replay.py +0 -0
  44. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/review_cli.py +0 -0
  45. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/review_queue.py +0 -0
  46. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/reviews.py +0 -0
  47. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/schemas.py +0 -0
  48. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/server.py +0 -0
  49. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/source_regions.py +0 -0
  50. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/sources.py +0 -0
  51. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/status.py +0 -0
  52. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/symbols.py +0 -0
  53. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/tex.py +0 -0
  54. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/util.py +0 -0
  55. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_AMS-Regular-BQhdFMY1.woff2 +0 -0
  56. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Caligraphic-Bold-Dq_IR9rO.woff2 +0 -0
  57. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Caligraphic-Regular-Di6jR-x-.woff2 +0 -0
  58. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Fraktur-Bold-CL6g_b3V.woff2 +0 -0
  59. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Fraktur-Regular-CTYiF6lA.woff2 +0 -0
  60. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Main-Bold-Cx986IdX.woff2 +0 -0
  61. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Main-BoldItalic-DxDJ3AOS.woff2 +0 -0
  62. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Main-Italic-NWA7e6Wa.woff2 +0 -0
  63. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Main-Regular-B22Nviop.woff2 +0 -0
  64. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Math-BoldItalic-CZnvNsCZ.woff2 +0 -0
  65. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Math-Italic-t53AETM-.woff2 +0 -0
  66. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_SansSerif-Bold-D1sUS0GD.woff2 +0 -0
  67. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_SansSerif-Italic-C3H0VqGB.woff2 +0 -0
  68. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_SansSerif-Regular-DDBCnlJ7.woff2 +0 -0
  69. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Script-Regular-D3wIWfF6.woff2 +0 -0
  70. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Size1-Regular-mCD8mA8B.woff2 +0 -0
  71. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Size2-Regular-Dy4dx90m.woff2 +0 -0
  72. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Size4-Regular-Dl5lxZxV.woff2 +0 -0
  73. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/KaTeX_Typewriter-Regular-CO6r4hn1.woff2 +0 -0
  74. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/cursor-editorTextHighlight-kG3uUOwL.svg +0 -0
  75. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/assets/pdf.worker.min-yatZIOMy.mjs +0 -0
  76. {machinechecked-0.1.15 → machinechecked-0.1.17}/engine/python/prooflens/web/theme.css +0 -0
  77. {machinechecked-0.1.15 → machinechecked-0.1.17}/setup.cfg +0 -0
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: machinechecked
3
- Version: 0.1.15
3
+ Version: 0.1.17
4
4
  Summary: Read a mathematical paper beside its Lean formalisation, and see claim by claim what Lean establishes.
5
5
  License: Apache-2.0
6
6
  Project-URL: Homepage, https://machinechecked.org
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: machinechecked
3
- Version: 0.1.15
3
+ Version: 0.1.17
4
4
  Summary: Read a mathematical paper beside its Lean formalisation, and see claim by claim what Lean establishes.
5
5
  License: Apache-2.0
6
6
  Project-URL: Homepage, https://machinechecked.org
@@ -40,6 +40,7 @@ engine/python/prooflens/status.py
40
40
  engine/python/prooflens/symbols.py
41
41
  engine/python/prooflens/tex.py
42
42
  engine/python/prooflens/util.py
43
+ engine/python/prooflens/versions.py
43
44
  engine/python/prooflens/data/ProofLensProbe.lean
44
45
  engine/python/prooflens/data/prooflens.sty
45
46
  engine/python/prooflens/data/statuses.json
@@ -69,6 +70,6 @@ engine/python/prooflens/web/assets/KaTeX_Size2-Regular-Dy4dx90m.woff2
69
70
  engine/python/prooflens/web/assets/KaTeX_Size4-Regular-Dl5lxZxV.woff2
70
71
  engine/python/prooflens/web/assets/KaTeX_Typewriter-Regular-CO6r4hn1.woff2
71
72
  engine/python/prooflens/web/assets/cursor-editorTextHighlight-kG3uUOwL.svg
72
- engine/python/prooflens/web/assets/index-B_fzeuBg.js
73
- engine/python/prooflens/web/assets/index-e4I0NErR.css
73
+ engine/python/prooflens/web/assets/index-B1CkNBtR.js
74
+ engine/python/prooflens/web/assets/index-DKOP2F1S.css
74
75
  engine/python/prooflens/web/assets/pdf.worker.min-yatZIOMy.mjs
@@ -1,7 +1,8 @@
1
1
  """The audit descriptor: `audit.json` at the root of a published reader.
2
2
 
3
3
  It says, in a few fields, what the audit is about and where its data is: the paper (and its
4
- arXiv identifier and version when it has one), the Lean repository and commit, who published
4
+ arXiv identifier when it has one), the version of the paper it pins (`paper_version`: published,
5
+ such as arXiv v1, or an unpublished manuscript at a commit), the Lean repository and commit, who published
5
6
  the audit, where reports go, and the path of the reader's data. A site that indexes audits reads
6
7
  this file and then the data it points to; it never runs anything an audit contains.
7
8
 
@@ -77,6 +78,12 @@ def descriptor(index: dict, lean_root: Path | None, author: str | None = None,
77
78
  entry.update(published_record(p["published"]))
78
79
  if p.get("published_note"):
79
80
  entry["note"] = p["published_note"]
81
+ # The version this audit pins (engine/python/prooflens/versions.py). An unpublished
82
+ # manuscript carries no published version number, even when `published` names the paper.
83
+ ver = p.get("paper_version")
84
+ entry["paper_version"] = ver
85
+ if ver and ver.get("kind") == "manuscript":
86
+ entry["version"] = None
80
87
  papers.append(entry)
81
88
  project = index.get("project") or {}
82
89
  return {
@@ -353,6 +353,9 @@ def _check_json(a):
353
353
  "links": [{"name": l["name"], "status": l["status"], "status_code": l["status_code"], "class": l["class"],
354
354
  "source": l.get("source")} for l in c["links"]],
355
355
  "correspondence": c.get("correspondence"),
356
+ "reviews": [{"reviewer": r["reviewer"], "verdict": r["verdict"]["value"],
357
+ "submitted_at": r["submitted_at"], "current": r["current"]}
358
+ for r in c.get("reviews", [])],
356
359
  "suggestions": [{"name": x["name"], "why": x.get("why")}
357
360
  for x in c.get("suggestions", [])]}
358
361
  for c in body["claims"]]
@@ -38,6 +38,10 @@ class Paper:
38
38
  # a DOI or a URL), and how the revision differs from it. Descriptive only: no status uses it.
39
39
  published: str | None = None
40
40
  published_note: str | None = None
41
+ # The version of the paper this audit pins: "published" (the version `published` names) or
42
+ # "manuscript" (the unpublished source at its pinned revision, `source.rev`). When absent, a
43
+ # paper with `published` (or an arXiv source) pins that published version (docs/ASSURANCE.md).
44
+ version: str | None = None
41
45
  # A paper whose source is a PDF: the PDF's SHA-256, pinned by the configuration. The release
42
46
  # gate accepts a PDF as source evidence only when it is pinned and matches (EVIDENCE_POLICY).
43
47
  pdf_sha256: str | None = None
@@ -118,11 +122,14 @@ def load(path: str | Path) -> Config:
118
122
  src = _req(p, "source", "paper")
119
123
  if not any(k in src for k in ("path", "git", "arxiv")):
120
124
  raise ConfigError(f"paper {p.get('id')}: source needs one of path, git, arxiv")
125
+ if p.get("version") not in (None, "published", "manuscript"):
126
+ raise ConfigError(f"paper {p.get('id')}: version must be \"published\" or "
127
+ f"\"manuscript\", not {p.get('version')!r}")
121
128
  papers.append(Paper(
122
129
  id=_req(p, "id", "paper"), title=p.get("title", p["id"]), source=src,
123
130
  root=p.get("root"), pdf=p.get("pdf"), pdf_sha256=p.get("pdf_sha256"), commit=src.get("rev"), envs=p.get("envs"),
124
131
  publish=Publish(**p.get("publish", {})), published=p.get("published"),
125
- published_note=p.get("note")))
132
+ published_note=p.get("note"), version=p.get("version")))
126
133
  return Config(
127
134
  path=path,
128
135
  name=_req(proj, "name", "project"),
@@ -282,7 +282,8 @@ def export(cfg: Config, out: Path, release: str | None = None,
282
282
  for p in index["papers"]:
283
283
  pub = pubs[p["id"]]
284
284
  q = _pick(p, ["id", "title", "mode", "revision", "pdf_sha256", "page_size", "note",
285
- "published", "published_note", "source_commit"])
285
+ "published", "published_note", "source_commit",
286
+ "paper_version"])
286
287
  q["root"] = p.get("root") if pub.source_text else None
287
288
  q["pdf"] = p["pdf"] if pub.pdf else None
288
289
  pidx["papers"].append(q)
@@ -142,8 +142,8 @@ def build(evidence: dict, release: str, display: dict | None = None
142
142
  "pdf": f"papers/{pid}.pdf" if prov.get("pdf_sha256") else None,
143
143
  "pdf_sha256": prov.get("pdf_sha256"), "page_size": t.get("page_size"),
144
144
  "note": prov.get("note", ""),
145
- **{k: prov[k] for k in ("published", "published_note", "source_commit")
146
- if prov.get(k)}})
145
+ **{k: prov[k] for k in ("published", "published_note", "source_commit",
146
+ "paper_version") if prov.get(k)}})
147
147
 
148
148
  store = decl_records(body, lean)
149
149
  keys = {n: record_key(r) for n, r in store.items()}
@@ -102,15 +102,62 @@ def _state(links: list[dict], audit: dict | None = None, audited: bool = False)
102
102
  return fields("link.not-in-audit" if audited else "link.no-audit")
103
103
 
104
104
 
105
+ HREF = re.compile(r"(?!//)(?![a-z][\w+.-]*:)[\w./~%-]*", re.I)
106
+ HTTPS = re.compile(r"https://[^\s\"'<>]+")
107
+
108
+
109
+ def page_versions(entries: list[dict], current: str) -> list[dict]:
110
+ """The versions of a paper its page offers (one page per version, each with its own PDF,
111
+ claim numbering and audits), as the reader's version drop-down receives them. Each entry
112
+ gives `id` and `kind` as an audit's `paper_version` does (a manuscript also its `revision`),
113
+ `href` (the address of that version's page, on the same site), and optionally `pdf` and
114
+ `audits` (`url`, `author`, `repository`). The label is MachineChecked's for that version,
115
+ never one the entry supplies; `current` is the id of the version this page shows."""
116
+ from .published import AuditError, paper_version
117
+ out = []
118
+ for e in entries:
119
+ try:
120
+ rec = paper_version({k: e.get(k) for k in ("kind", "id", "revision")})
121
+ except AuditError as err:
122
+ raise ValueError(f"paper version {e.get('id')!r}: {err}") from None
123
+ href = e.get("href")
124
+ if not isinstance(href, str) or not href or not HREF.fullmatch(href):
125
+ raise ValueError(f"paper version {rec['id']}: href must be an address on this site")
126
+ pdf = e.get("pdf")
127
+ if pdf is not None and not (isinstance(pdf, str) and (HTTPS.fullmatch(pdf) or
128
+ HREF.fullmatch(pdf))):
129
+ raise ValueError(f"paper version {rec['id']}: pdf must be https or on this site")
130
+ audits = [{k: a.get(k) for k in ("url", "author", "repository") if a.get(k)}
131
+ for a in e.get("audits") or [] if HTTPS.fullmatch(str(a.get("url", "")))]
132
+ out.append({"id": rec["id"], "kind": rec["kind"], "label": rec["label"],
133
+ **({"revision": rec["revision"]} if rec.get("revision") else {}),
134
+ "href": href, "pdf": pdf, "audits": audits})
135
+ ids = [v["id"] for v in out]
136
+ if len(set(ids)) != len(ids) or ids.count(current) != 1:
137
+ raise ValueError(f"the versions must be distinct and include this page's ({current})")
138
+ return out
139
+
140
+
105
141
  def write_paper_site(site: Path, paper: dict, claims: list[dict],
106
142
  page_size: list | tuple | None, pdf_url: str | None,
107
- web_dist: Path | None = None, audits: dict | None = None) -> dict:
143
+ web_dist: Path | None = None, audits: dict | None = None,
144
+ versions: list[dict] | None = None, version: str | None = None) -> dict:
108
145
  """Write the paper page's data under `site/data/` (and the reader, given its built
109
146
  directory). `paper` is the paper's registry entry (its identifier, title, authors, links and
110
147
  formalisations), which the reader's panel beside the paper shows; `claims` are from
111
- `paper_claims`. Replaces whatever is at `site`."""
148
+ `paper_claims`. A paper with several versions has a page per version: `version` is the id of
149
+ this page's, `versions` every version's entry (`page_versions`), and every audit loaded for
150
+ this page must pin this version. Replaces whatever is at `site`."""
112
151
  from .pipeline import write_preview
152
+ from .published import same_version
113
153
  audits_loaded = audits or {}
154
+ offered = page_versions(versions, version) if versions else None
155
+ if offered:
156
+ for r, a in audits_loaded.items():
157
+ pv = a.get("paper_version") if a.get("state") == "loaded" else None
158
+ if a.get("state") == "loaded" and not (pv and same_version(pv["id"], version)):
159
+ raise ValueError(f"the audit of {r} does not pin {version}: load it with the "
160
+ "listing's version (load_published_audit)")
114
161
  projects = [f["repository"] for f in paper.get("formalisations", [])]
115
162
  audited = {r for r, a in audits_loaded.items() if a.get("state") == "loaded"} if audits is not None else {f["repository"] for f in paper.get("formalisations", []) if f.get("audits")}
116
163
  index_claims, chunks = [], {}
@@ -167,10 +214,12 @@ def write_paper_site(site: Path, paper: dict, claims: list[dict],
167
214
  editions = {a["paper"]["edition"]: {**a["paper"], "author": a.get("author"), "repository": r}
168
215
  for r, a in audits_loaded.items() if a.get("paper", {}).get("edition")}
169
216
  index = {"schema": SCHEMA, "editions": list(editions.values()),
170
- "audits": {r: {k: a[k] for k in ("state", "detail", "commit", "release", "author", "url", "gates") if k in a}
217
+ "audits": {r: {k: a[k] for k in ("state", "detail", "commit", "release", "author", "url", "gates",
218
+ "paper_version") if k in a}
171
219
  for r, a in audits_loaded.items()}, "mode": "paper", "release": "paper", "evidence": eid,
172
220
  "tool": {"name": "machinechecked"}, "project": {}, "projects": projects,
173
221
  "paper": paper, "notes": [],
222
+ **({"version": version, "versions": offered} if offered else {}),
174
223
  "papers": [{"id": "P", "title": paper.get("title", ""), "mode": "pdf", "pdf": None,
175
224
  "pdf_url": pdf_url, "page_size": list(page_size) if page_size else None,
176
225
  "note": ""}],
@@ -20,6 +20,7 @@ from .schemas import validate
20
20
  from .sources import acquire
21
21
  from .source_regions import collect as collect_source_regions
22
22
  from .tex import DEFAULT_ENVS, TexError, anchors_for, theorem_envs
23
+ from . import versions as paper_versions
23
24
  from .util import (canonical_json, git_dirty, git_head, link_or_copy, prepare_work_dir,
24
25
  read_json, sha256_file, sha256_text, write_json)
25
26
 
@@ -62,6 +63,11 @@ def build(cfg: Config, release: str | None = None, skip_lean: bool = False,
62
63
  # runs: a problem there should not wait for a probe that can take hours.
63
64
  snaps = {paper.id: acquire(cfg, paper, cfg.out_dir / "sources" / paper.id)
64
65
  for paper in cfg.papers}
66
+ # The version of each paper the audit pins, refused when the inventory pins another one.
67
+ versions = {}
68
+ for paper in cfg.papers:
69
+ rec, is_declared = paper_versions.declared(paper, snaps[paper.id], cfg.root)
70
+ versions[paper.id] = paper_versions.check(paper, rec, is_declared, inv)
65
71
 
66
72
  lean = None
67
73
  lean_digest = None
@@ -94,7 +100,8 @@ def build(cfg: Config, release: str | None = None, skip_lean: bool = False,
94
100
  # A paper read from a folder: the commit of the repository that folder is in.
95
101
  **({"source_commit": git_head(cfg.root / paper.source["path"])}
96
102
  if snap.kind == "path" and git_head(cfg.root / paper.source["path"]) else {}),
97
- **({"published_note": paper.published_note} if paper.published_note else {})}
103
+ **({"published_note": paper.published_note} if paper.published_note else {}),
104
+ **({"paper_version": versions[paper.id]} if versions[paper.id] else {})}
98
105
  t = None
99
106
  if snap.mode == "latex" and not skip_tex:
100
107
  envs = paper.envs or sorted(set(DEFAULT_ENVS) | set(theorem_envs(snap.dir)))
@@ -10,6 +10,7 @@ import re
10
10
  from urllib.parse import quote, urlsplit, urljoin
11
11
 
12
12
  from . import manifest as mf
13
+ from . import versions
13
14
  from .status import fields, STATUSES
14
15
  from .util import content_id
15
16
  from .schemas import validate
@@ -34,6 +35,39 @@ def audit_version(desc: dict) -> str:
34
35
  return content_id({k: desc.get(k) for k in ('schema', 'release', 'formalisation', 'papers', 'data')})
35
36
 
36
37
 
38
+ def paper_version(value) -> dict | None:
39
+ """The paper version an audit's descriptor says it pins, rebuilt from its kind and
40
+ identifier: the label shown is the one MachineChecked gives that version, never a label the
41
+ audit supplies (a manuscript is never shown with a version number). `checked` is true only
42
+ when the audit says every claim's pin was checked against the version; an audit that does
43
+ not say so is read as unchecked."""
44
+ if value is None:
45
+ return None
46
+ if not isinstance(value, dict):
47
+ raise AuditError('Invalid paper version')
48
+ kind, vid, rev = value.get('kind'), value.get('id'), value.get('revision')
49
+ checked = value.get('checked') is True
50
+ if kind == 'manuscript':
51
+ if not isinstance(rev, str) or not versions.COMMIT.fullmatch(rev) or \
52
+ vid != 'manuscript:' + rev[:versions.SHORT]:
53
+ raise AuditError('Invalid manuscript version')
54
+ return {**versions.manuscript_record(rev), 'checked': checked}
55
+ if kind == 'published' and isinstance(vid, str) and vid:
56
+ rec = versions.published_version(('arXiv:' + vid[6:]) if vid.startswith('arxiv:') else vid)
57
+ if rec['id'] != vid:
58
+ raise AuditError('Invalid published version')
59
+ return {**rec, 'revision': rev if isinstance(rev, str) else None, 'checked': checked}
60
+ raise AuditError('Invalid paper version')
61
+
62
+
63
+ def same_version(a: str, b: str) -> bool:
64
+ """Whether two version identifiers name one version (a manuscript's commit may be given
65
+ abbreviated)."""
66
+ if a.startswith('manuscript:') and b.startswith('manuscript:'):
67
+ return versions.same_commit(a[11:], b[11:])
68
+ return a.lower() == b.lower()
69
+
70
+
37
71
  def load_published_audit(paper: dict, project: dict, listing: dict, fetch) -> dict:
38
72
  """Fetch and validate one snapshot. ``fetch(url, byte_limit)`` returns bytes.
39
73
 
@@ -87,6 +121,11 @@ def load_published_audit(paper: dict, project: dict, listing: dict, fetch) -> di
87
121
  expected_version = project.get('paper_version')
88
122
  if expected_version and public.get('version') != expected_version:
89
123
  raise AuditError('Audit paper version differs from the listing')
124
+ pinned = paper_version(public.get('paper_version'))
125
+ # A listing that names the version it lists this audit under (`version`, an identifier as
126
+ # `paper_version.id` gives it) needs an audit that pins that version.
127
+ if listing.get('version') and not (pinned and same_version(listing['version'], pinned['id'])):
128
+ raise AuditError('Audit pins another version of the paper than the listing names')
90
129
  raw_index = read(d['data'], 20_000_000)
91
130
  idx = deepcopy(raw_index)
92
131
  if idx.get('schema') != mf.SCHEMA:
@@ -106,6 +145,8 @@ def load_published_audit(paper: dict, project: dict, listing: dict, fetch) -> di
106
145
  raise AuditError('Audit needs an unambiguous internal paper key')
107
146
  ap = matched[0]
108
147
  revision = ap.get('revision') or ap.get('source_commit')
148
+ if paper_version(ap.get('paper_version')) != pinned:
149
+ raise AuditError('Audit descriptor and data name different paper versions')
109
150
  if not revision or revision != public.get('revision'):
110
151
  raise AuditError('Audit paper revisions disagree')
111
152
  rows = [c for c in idx.get('claims', []) if c.get('paper') == ap['id']]
@@ -157,6 +198,7 @@ def load_published_audit(paper: dict, project: dict, listing: dict, fetch) -> di
157
198
  raise AuditError('Audit changed during download; retry the build')
158
199
  return {'author': d.get('author') or listing.get('author'), 'url': base,
159
200
  'commit': f['commit'], 'release': idx['release'], 'version': audit_version(d),
201
+ 'paper_version': pinned,
160
202
  'claims': records, 'store': store, 'pdfs': pdfs, 'paper': audited_paper,
161
203
  'context': {k: idx.get(k) for k in ('release', 'evidence', 'project', 'lean', 'issues', 'site_url', 'created', 'tool')},
162
204
  'gates': idx.get('gates', {}).get(ap['id']), 'state': 'loaded'}
@@ -261,7 +261,7 @@ def linked_names(claim: dict, entry: dict) -> list[str]:
261
261
  return [n for n in names if n not in entry.get("rejected", [])]
262
262
 
263
263
 
264
- COVERAGE = ("full", "partial", "changed", "related")
264
+ COVERAGE = ("full", "composed", "partial", "changed", "related", "unresolved")
265
265
  CONFIDENCE = ("high", "medium", "low")
266
266
  NOTE_CHARS = 2000
267
267
 
@@ -288,6 +288,71 @@ def correspondence(claim: dict, entry: dict) -> dict | None:
288
288
  "note": " ".join(notes)[:NOTE_CHARS], "by": None}
289
289
 
290
290
 
291
+ REVIEWS_DIR = "machinechecked-reviews"
292
+ REVIEW_TEXT_CHARS = 20000
293
+
294
+
295
+ def load_review_reports(root: Path) -> tuple[dict[str, list[dict]], list[str]]:
296
+ """The review reports exported by the reader's Review page and kept in the project's
297
+ `machinechecked-reviews/` folder, by review unit (`link:<claim key>`), and a note for each
298
+ file that could not be read. A report's text is the reviewer's and is shown as text."""
299
+ out: dict[str, list[dict]] = {}
300
+ notes: list[str] = []
301
+ seen = set()
302
+ d = root / REVIEWS_DIR
303
+ for f in sorted(d.glob("*.json")) if d.is_dir() else []:
304
+ try:
305
+ data = json.loads(f.read_text(encoding="utf-8"))
306
+ reports = data["reports"] if data.get("schema") == "prooflens/review-reports@1" else None
307
+ if not isinstance(reports, list):
308
+ raise ValueError("not a review export")
309
+ except (OSError, ValueError, KeyError, AttributeError) as e:
310
+ notes.append(f"{REVIEWS_DIR}/{f.name} was skipped: {e}")
311
+ continue
312
+ for r in reports:
313
+ if not isinstance(r, dict) or r.get("schema") != "prooflens/review-report@1":
314
+ continue
315
+ unit = (r.get("unit") or {}).get("id")
316
+ key = (unit, (r.get("reviewer") or {}).get("id"), r.get("submitted_at"))
317
+ if not isinstance(unit, str) or key in seen:
318
+ continue
319
+ seen.add(key)
320
+ out.setdefault(unit, []).append(r)
321
+ return out, notes
322
+
323
+
324
+ def _text(x) -> str:
325
+ return x[:REVIEW_TEXT_CHARS] if isinstance(x, str) else ""
326
+
327
+
328
+ def review_view(r: dict, links: list[dict]) -> dict:
329
+ """A report as the reader shows it, and whether it is current: it reviewed exactly the
330
+ claim's linked declarations, with the statements they have now (`None` when the Lean was
331
+ not checked, so their statements are unknown)."""
332
+ pins = r.get("pins") or {}
333
+ fps = pins.get("fingerprints") or {}
334
+ now = {}
335
+ for l in links:
336
+ rec = l.get("record") or {}
337
+ now[l["name"]] = rec.get("fingerprint") or (rec.get("decl") or {}).get("fingerprint")
338
+ if not links or any(v is None for v in now.values()):
339
+ current = None
340
+ else:
341
+ current = set(fps) == set(now) and all(fps[n] == now[n] for n in now)
342
+ rv, verdict = r.get("reviewer") or {}, r.get("verdict") or {}
343
+ return {"reviewer": {k: rv.get(k) for k in ("id", "kind", "name", "model", "identity",
344
+ "same_family_as_author")},
345
+ "verdict": {"group": verdict.get("group"), "value": verdict.get("value"),
346
+ "detail": _text(verdict.get("detail"))},
347
+ "reading": _text((r.get("reading") or {}).get("text")),
348
+ "comparison": _text((r.get("comparison") or {}).get("text")),
349
+ "checklist": (r.get("comparison") or {}).get("checklist") or {},
350
+ "locked_at": (r.get("reading") or {}).get("locked_at"),
351
+ "submitted_at": r.get("submitted_at"),
352
+ "reviewed": sorted(fps), "release": pins.get("release"),
353
+ "project_commit": pins.get("project_commit"), "current": current}
354
+
355
+
291
356
  def link_source(claim: dict, entry: dict, name: str) -> str:
292
357
  if "lean" in entry:
293
358
  return "confirmed"
@@ -644,6 +709,8 @@ def build_view(root: Path, paper_path: Path, out_dir: Path, lake: str = "lake",
644
709
  if st.get("module")}
645
710
  c["declared_notes"] = [st["note"] for st in rows if st.get("note")]
646
711
  links = load_links(root).get("papers", {}).get(paper_key, {})
712
+ reports, review_notes = load_review_reports(root)
713
+ notes += review_notes
647
714
  if not libs:
648
715
  notes.append("no Lean library found in the lakefile")
649
716
  elif not built:
@@ -741,6 +808,7 @@ def build_view(root: Path, paper_path: Path, out_dir: Path, lake: str = "lake",
741
808
  found = [r for r in linkrecs.get(s["name"], []) if "missing" not in r.get("errors", [])]
742
809
  s["record"] = found[0] if len({((r.get("decl") or {}).get("module")) for r in found}) == 1 else None
743
810
  c["correspondence"] = correspondence(c, entry) if c["links"] else None
811
+ c["reviews"] = [review_view(r, c["links"]) for r in reports.get("link:" + c["key"], [])]
744
812
  if c["links"]:
745
813
  worst = max(c["links"], key=lambda l: ["proved", "unspecified", "open", "malformed"]
746
814
  .index(l["class"]))
@@ -830,7 +898,7 @@ def write_view_site(view: dict, site: Path, web_dist: Path | None,
830
898
  "source_evidence": source_evidence({"kind": akind} if anchor else None, body["tex"]["mode"], paper=body["tex"]),
831
899
  "blockers": [],
832
900
  "links": [_link_view(l, root) for l in c["links"]],
833
- "correspondence": c.get("correspondence"),
901
+ "correspondence": c.get("correspondence"), "reviews": c.get("reviews", []),
834
902
  "suggestions": [_link_view(s, root) for s in c.get("suggestions", [])],
835
903
  "clauses": [], "decls": {n: keys[n] for n in sorted(names) if n in keys},
836
904
  "findings": []}
@@ -849,6 +917,7 @@ def write_view_site(view: dict, site: Path, web_dist: Path | None,
849
917
  "lean": None, "notes": body["notes"],
850
918
  "papers": [{"id": "P", "title": body["title"], "mode": body["tex"]["mode"],
851
919
  "revision": body.get("commit"),
920
+ **_paper_version(body),
852
921
  "pdf": "papers/P.pdf" if view["pdf"] else None,
853
922
  "pdf_sha256": body.get("pdf_sha256") if view["pdf"] else None,
854
923
  "page_size": body["tex"].get("page_size"), "note": ""}],
@@ -873,6 +942,23 @@ def write_view_site(view: dict, site: Path, web_dist: Path | None,
873
942
  return index
874
943
 
875
944
 
945
+ def _paper_version(body: dict) -> dict:
946
+ """The version of the paper the view reads: the arXiv version its URL names, else the
947
+ project's commit (the paper is a file of the project). Its claims are read from that
948
+ version itself, so the version is checked."""
949
+ from . import versions
950
+ from .audit import arxiv_id
951
+ key = body.get("paper") or ""
952
+ ax = arxiv_id(key) if key.startswith(("http://", "https://")) else None
953
+ if ax and ax[1]:
954
+ return {"paper_version": {**versions.published_version(f"arXiv:{ax[0][6:]}{ax[1]}"),
955
+ "checked": True}}
956
+ commit = body.get("commit") or ""
957
+ if versions.COMMIT.fullmatch(commit) and len(commit) == 40:
958
+ return {"paper_version": {**versions.manuscript_record(commit), "checked": True}}
959
+ return {}
960
+
961
+
876
962
  def _link_view(l: dict, root: Path | None = None) -> dict:
877
963
  """A link or suggestion as the reader shows it; `path` is its file in the project's
878
964
  repository, for a link to its source, when the project is known."""
@@ -0,0 +1,171 @@
1
+ """The version of a paper an audit pins, and the check that the claim inventory pins the same one.
2
+
3
+ A version is either published (the version `published` names, such as `arXiv:2601.01234v1`, or
4
+ an arXiv source) or an unpublished manuscript at a commit (`version = "manuscript"`, with the
5
+ source's pinned revision). Its record says which, with a label for readers: "arXiv v1", or
6
+ "Latest manuscript (commit abc1234)". An unpublished manuscript is never given a version number.
7
+
8
+ The record is descriptive: no status depends on it. What is checked is that the inventory's pins
9
+ (`papers.<id>.commit`, a claim's `commit`) name the declared version: a published version by its
10
+ identifier (`2601.01234v1`), a manuscript by its commit. So an audit labelled with one version
11
+ cannot carry claims pinned to another. The record's `checked` says whether that check covers
12
+ every claim: it is false when some claim carries no pin, so the version is only as the
13
+ configuration states it, and readers are told so.
14
+ """
15
+
16
+ from __future__ import annotations
17
+
18
+ import re
19
+ import subprocess
20
+ from pathlib import Path
21
+
22
+ from .audit import published_record
23
+
24
+ COMMIT = re.compile(r"[0-9a-f]{7,40}")
25
+ ARXIV_PIN = re.compile(r"(?:arxiv:)?([a-z-]+(?:\.[A-Z]{2})?/\d{7}|\d{4}\.\d{4,5})(v\d+)?", re.I)
26
+ SHORT = 7
27
+
28
+
29
+ class VersionError(ValueError):
30
+ """A paper's declared version is incomplete or disagrees with the inventory's pins."""
31
+
32
+
33
+ def same_commit(a: str, b: str) -> bool:
34
+ """Whether two commit names (full or abbreviated, at least 7 hexadecimal digits) agree."""
35
+ a, b = a.strip().lower(), b.strip().lower()
36
+ return a == b or bool(COMMIT.fullmatch(a) and COMMIT.fullmatch(b)
37
+ and (a.startswith(b) or b.startswith(a)))
38
+
39
+
40
+ def _toplevel_head(folder: Path) -> str | None:
41
+ """The commit checked out in `folder`, when `folder` is the top of its own Git work tree
42
+ (a folder inside another repository is not the manuscript's own history)."""
43
+ if not folder.is_dir():
44
+ return None
45
+ top = subprocess.run(["git", "rev-parse", "--show-toplevel"], cwd=folder,
46
+ capture_output=True, text=True)
47
+ if top.returncode != 0 or Path(top.stdout.strip()).resolve() != folder.resolve():
48
+ return None
49
+ head = subprocess.run(["git", "rev-parse", "HEAD"], cwd=folder, capture_output=True,
50
+ text=True)
51
+ return head.stdout.strip() if head.returncode == 0 else None
52
+
53
+
54
+ def manuscript_revision(paper, snap, base: Path) -> str | None:
55
+ """The commit of the source the audit reads: a Git source's resolved `rev`; a folder's
56
+ stated `rev` (checked against the folder's own history when it has one), else the folder's
57
+ commit when it is a repository of its own."""
58
+ if snap.kind == "git":
59
+ return snap.revision
60
+ if snap.kind != "path":
61
+ return None
62
+ stated = paper.source.get("rev")
63
+ head = _toplevel_head((base / paper.source["path"]).resolve())
64
+ if stated and head and not same_commit(stated, head):
65
+ raise VersionError(f"paper {paper.id}: source rev {stated} is not the commit checked "
66
+ f"out in {paper.source['path']} ({head[:SHORT]})")
67
+ return str(stated) if stated else head
68
+
69
+
70
+ def manuscript_record(revision: str) -> dict:
71
+ short = revision[:SHORT]
72
+ return {"kind": "manuscript", "id": f"manuscript:{short}", "revision": revision,
73
+ "label": f"Latest manuscript (commit {short})"}
74
+
75
+
76
+ def published_version(value: str) -> dict:
77
+ """The record of a published version named as `published` names it."""
78
+ rec = published_record(value)
79
+ if rec["id"] and rec["id"].startswith("arxiv:"):
80
+ n = rec["version"]
81
+ return {"kind": "published", "id": rec["id"] + (n or ""), "arxiv": rec["id"],
82
+ "number": n, "label": f"arXiv {n}" if n else f"arXiv {rec['id'][6:]}"}
83
+ if rec["id"]:
84
+ return {"kind": "published", "id": rec["id"], "label": f"Published ({rec['id']})"}
85
+ return {"kind": "published", "id": rec["url"] or value, "label": "Published version"}
86
+
87
+
88
+ def declared(paper, snap, base: Path) -> tuple[dict | None, bool]:
89
+ """The version this paper's audit pins, and whether the configuration declares it (a
90
+ declared version is checked against the inventory; an inferred one is dropped instead)."""
91
+ revision = manuscript_revision(paper, snap, base)
92
+ want = paper.version
93
+ source_arxiv = snap.revision if snap.kind == "arxiv" else None
94
+ if want == "published" or (want is None and (paper.published or source_arxiv)):
95
+ if source_arxiv:
96
+ rec = published_version("arXiv:" + source_arxiv)
97
+ if paper.published and published_version(paper.published)["id"] != rec["id"]:
98
+ raise VersionError(f"paper {paper.id}: published names {paper.published}, but "
99
+ f"the source is arXiv {source_arxiv}")
100
+ elif paper.published:
101
+ rec = published_version(paper.published)
102
+ else:
103
+ raise VersionError(f'paper {paper.id}: version = "published" needs `published`, '
104
+ 'as in published = "arXiv:2601.01234v1"')
105
+ return {**rec, "revision": revision}, True
106
+ if want == "manuscript":
107
+ if snap.kind == "arxiv":
108
+ raise VersionError(f'paper {paper.id}: version = "manuscript", but the source is '
109
+ 'arXiv; read the manuscript from its repository (git, rev)')
110
+ if not revision:
111
+ raise VersionError(f'paper {paper.id}: version = "manuscript" needs the commit of '
112
+ 'the manuscript: give the source a `rev`')
113
+ return manuscript_record(revision), True
114
+ return (manuscript_record(revision) if revision else None), False
115
+
116
+
117
+ def inventory_pins(inv: dict, pid: str) -> dict[str, list[str]]:
118
+ """Each pin of the paper's claims (a claim's `commit`, else `papers.<id>.commit`), with the
119
+ claims it pins."""
120
+ default = ((inv.get("papers") or {}).get(pid) or {}).get("commit")
121
+ out: dict[str, list[str]] = {}
122
+ for c in inv.get("claims", []):
123
+ if c.get("paper") == pid and (c.get("commit") or default):
124
+ out.setdefault(str(c.get("commit") or default), []).append(c["id"])
125
+ return out
126
+
127
+
128
+ def pin_matches(pin: str, rec: dict) -> bool:
129
+ """Whether an inventory pin names the version `rec` records. A published version is named
130
+ by its own identifier (`2601.01234v1`, with or without `arXiv:`); a commit cannot show
131
+ which published version it is. A manuscript is named by its commit."""
132
+ pin = pin.strip()
133
+ if rec["kind"] == "published":
134
+ m = ARXIV_PIN.fullmatch(pin)
135
+ if m and not COMMIT.fullmatch(pin.lower()):
136
+ return ("arxiv:" + m.group(1) + (m.group(2) or "")).lower() == rec["id"].lower()
137
+ return pin.lower() == rec["id"].lower()
138
+ return bool(COMMIT.fullmatch(pin.lower()) and same_commit(pin, rec["revision"]))
139
+
140
+
141
+ def describe_pin(pin: str) -> str:
142
+ return f"commit {pin[:SHORT]}" if COMMIT.fullmatch(pin.lower()) else pin
143
+
144
+
145
+ def unpinned_claims(inv: dict, pid: str) -> list[str]:
146
+ """The paper's claims that carry no pin (neither their own `commit` nor the paper's)."""
147
+ if ((inv.get("papers") or {}).get(pid) or {}).get("commit"):
148
+ return []
149
+ return [c["id"] for c in inv.get("claims", []) if c.get("paper") == pid and not c.get("commit")]
150
+
151
+
152
+ def check(paper, rec: dict | None, is_declared: bool, inv: dict) -> dict | None:
153
+ """The version record to publish: `rec` when the inventory's pins agree with it, with
154
+ `checked` false when some claim carries no pin to check it against (or there are no claims).
155
+ A declared version the pins disagree with is refused; an inferred one is dropped (nothing is
156
+ claimed)."""
157
+ if rec is None:
158
+ return None
159
+ pins = inventory_pins(inv, paper.id)
160
+ bad = {pin: ids for pin, ids in pins.items() if not pin_matches(pin, rec)}
161
+ if not bad:
162
+ return {**rec, "checked": bool(pins) and not unpinned_claims(inv, paper.id)}
163
+ if not is_declared:
164
+ return None
165
+ pins = "; ".join(f"{describe_pin(p)} ({len(ids)} claim{'s' * (len(ids) != 1)}, "
166
+ f"such as {ids[0]})" for p, ids in sorted(bad.items()))
167
+ fix = ('pin the inventory to ' + (rec['id'].removeprefix('arxiv:') if rec['kind'] ==
168
+ 'published' else f"commit {rec['revision'][:SHORT]}")
169
+ + ', or declare the version the inventory pins (`version` and the source\'s `rev`)')
170
+ raise VersionError(f"paper {paper.id}: the audit pins {rec['label']}, but its inventory is "
171
+ f"pinned to {pins}. An audit pins one version of its paper: {fix}.")