machinechecked 0.1.14__tar.gz → 0.1.16__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 (78) hide show
  1. {machinechecked-0.1.14 → machinechecked-0.1.16}/PKG-INFO +1 -1
  2. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/machinechecked.egg-info/PKG-INFO +1 -1
  3. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/machinechecked.egg-info/SOURCES.txt +4 -2
  4. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/audit.py +8 -1
  5. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/cache.py +20 -8
  6. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/cli.py +1 -0
  7. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/config.py +8 -1
  8. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/data/ProofLensProbe.lean +23 -9
  9. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/export.py +6 -2
  10. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/formalization.py +2 -1
  11. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/manifest.py +2 -2
  12. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/paper.py +52 -3
  13. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/pipeline.py +17 -7
  14. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/published.py +42 -0
  15. machinechecked-0.1.16/engine/python/prooflens/repin.py +473 -0
  16. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/simple.py +56 -1
  17. machinechecked-0.1.16/engine/python/prooflens/versions.py +171 -0
  18. machinechecked-0.1.14/engine/python/prooflens/web/assets/index-GfZV3aEH.css → machinechecked-0.1.16/engine/python/prooflens/web/assets/index-BNMmM1YT.css +1 -1
  19. machinechecked-0.1.16/engine/python/prooflens/web/assets/index-CRqix-cK.js +278 -0
  20. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/index.html +2 -2
  21. {machinechecked-0.1.14 → machinechecked-0.1.16}/pyproject.toml +1 -1
  22. machinechecked-0.1.14/engine/python/prooflens/web/assets/index-YAI0NIYT.js +0 -278
  23. {machinechecked-0.1.14 → machinechecked-0.1.16}/LICENSE +0 -0
  24. {machinechecked-0.1.14 → machinechecked-0.1.16}/README.md +0 -0
  25. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/machinechecked.egg-info/dependency_links.txt +0 -0
  26. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/machinechecked.egg-info/entry_points.txt +0 -0
  27. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/machinechecked.egg-info/requires.txt +0 -0
  28. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/machinechecked.egg-info/top_level.txt +0 -0
  29. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/__init__.py +0 -0
  30. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/__main__.py +0 -0
  31. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/assess.py +0 -0
  32. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/bundle.py +0 -0
  33. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/data/prooflens.sty +0 -0
  34. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/data/schema/evidence.schema.json +0 -0
  35. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/data/schema/inventory.schema.json +0 -0
  36. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/data/schema/manifest-claim.schema.json +0 -0
  37. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/data/schema/manifest-index.schema.json +0 -0
  38. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/data/statuses.json +0 -0
  39. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/deps.py +0 -0
  40. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/evidence.py +0 -0
  41. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/pdfclaims.py +0 -0
  42. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/probe.py +0 -0
  43. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/promote.py +0 -0
  44. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/replay.py +0 -0
  45. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/review_cli.py +0 -0
  46. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/review_queue.py +0 -0
  47. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/reviews.py +0 -0
  48. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/schemas.py +0 -0
  49. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/server.py +0 -0
  50. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/source_regions.py +0 -0
  51. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/sources.py +0 -0
  52. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/status.py +0 -0
  53. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/symbols.py +0 -0
  54. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/tex.py +0 -0
  55. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/util.py +0 -0
  56. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_AMS-Regular-BQhdFMY1.woff2 +0 -0
  57. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Caligraphic-Bold-Dq_IR9rO.woff2 +0 -0
  58. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Caligraphic-Regular-Di6jR-x-.woff2 +0 -0
  59. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Fraktur-Bold-CL6g_b3V.woff2 +0 -0
  60. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Fraktur-Regular-CTYiF6lA.woff2 +0 -0
  61. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Main-Bold-Cx986IdX.woff2 +0 -0
  62. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Main-BoldItalic-DxDJ3AOS.woff2 +0 -0
  63. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Main-Italic-NWA7e6Wa.woff2 +0 -0
  64. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Main-Regular-B22Nviop.woff2 +0 -0
  65. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Math-BoldItalic-CZnvNsCZ.woff2 +0 -0
  66. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Math-Italic-t53AETM-.woff2 +0 -0
  67. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_SansSerif-Bold-D1sUS0GD.woff2 +0 -0
  68. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_SansSerif-Italic-C3H0VqGB.woff2 +0 -0
  69. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_SansSerif-Regular-DDBCnlJ7.woff2 +0 -0
  70. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Script-Regular-D3wIWfF6.woff2 +0 -0
  71. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Size1-Regular-mCD8mA8B.woff2 +0 -0
  72. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Size2-Regular-Dy4dx90m.woff2 +0 -0
  73. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Size4-Regular-Dl5lxZxV.woff2 +0 -0
  74. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/KaTeX_Typewriter-Regular-CO6r4hn1.woff2 +0 -0
  75. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/cursor-editorTextHighlight-kG3uUOwL.svg +0 -0
  76. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/assets/pdf.worker.min-yatZIOMy.mjs +0 -0
  77. {machinechecked-0.1.14 → machinechecked-0.1.16}/engine/python/prooflens/web/theme.css +0 -0
  78. {machinechecked-0.1.14 → machinechecked-0.1.16}/setup.cfg +0 -0
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: machinechecked
3
- Version: 0.1.14
3
+ Version: 0.1.16
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: 0.1.14
3
+ Version: 0.1.16
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
@@ -26,6 +26,7 @@ engine/python/prooflens/pipeline.py
26
26
  engine/python/prooflens/probe.py
27
27
  engine/python/prooflens/promote.py
28
28
  engine/python/prooflens/published.py
29
+ engine/python/prooflens/repin.py
29
30
  engine/python/prooflens/replay.py
30
31
  engine/python/prooflens/review_cli.py
31
32
  engine/python/prooflens/review_queue.py
@@ -39,6 +40,7 @@ engine/python/prooflens/status.py
39
40
  engine/python/prooflens/symbols.py
40
41
  engine/python/prooflens/tex.py
41
42
  engine/python/prooflens/util.py
43
+ engine/python/prooflens/versions.py
42
44
  engine/python/prooflens/data/ProofLensProbe.lean
43
45
  engine/python/prooflens/data/prooflens.sty
44
46
  engine/python/prooflens/data/statuses.json
@@ -68,6 +70,6 @@ engine/python/prooflens/web/assets/KaTeX_Size2-Regular-Dy4dx90m.woff2
68
70
  engine/python/prooflens/web/assets/KaTeX_Size4-Regular-Dl5lxZxV.woff2
69
71
  engine/python/prooflens/web/assets/KaTeX_Typewriter-Regular-CO6r4hn1.woff2
70
72
  engine/python/prooflens/web/assets/cursor-editorTextHighlight-kG3uUOwL.svg
71
- engine/python/prooflens/web/assets/index-GfZV3aEH.css
72
- engine/python/prooflens/web/assets/index-YAI0NIYT.js
73
+ engine/python/prooflens/web/assets/index-BNMmM1YT.css
74
+ engine/python/prooflens/web/assets/index-CRqix-cK.js
73
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 {
@@ -220,13 +220,18 @@ def check_space(project: Path, tmp: Path, man: dict) -> None:
220
220
  "revision, or old build output) and restore again; nothing was changed")
221
221
 
222
222
 
223
- def _unpack(arc: Path, project: Path) -> None:
223
+ def _unpack(arc: Path, project: Path, older: bool = False) -> None:
224
224
  """Unpack an archive written by `_pack` into the project and its path dependencies, with the
225
225
  tar `data` filter (no absolute paths, no `..`, no links out of the destination). Everything
226
226
  is unpacked into a staging folder first and moved into place only once the whole archive
227
- has been read, so a failure (a full disk, a corrupt archive) leaves the build as it was."""
227
+ has been read, so a failure (a full disk, a corrupt archive) leaves the build as it was.
228
+
229
+ An archive in the older layout (`older`) holds paths from the checkout's parent, under the
230
+ folder names of the checkout it was made in; each top folder is mapped onto the project or a
231
+ path dependency whose folder has that name in any case, and any other path is refused."""
228
232
  import tarfile
229
233
  deps = {d.name: d for d in path_deps(project)}
234
+ folded = {d.name.lower(): d for d in deps.values()}
230
235
  stage = Path(tempfile.mkdtemp(prefix=".lean-cache-restore-", dir=project))
231
236
  try:
232
237
  bases: dict[str, Path] = {}
@@ -234,7 +239,17 @@ def _unpack(arc: Path, project: Path) -> None:
234
239
  with tarfile.open(fileobj=z.stdout, mode="r|") as tf:
235
240
  for m in tf:
236
241
  head, _, rest = m.name.partition("/")
237
- if head == "project" and rest:
242
+ if older and not rest.strip("/") and m.isdir():
243
+ continue # the top folder itself
244
+ if older and rest and head.lower() == project.name.lower():
245
+ key, base, m.name = "project", project, rest
246
+ elif older and rest and head.lower() in folded:
247
+ base = folded[head.lower()]
248
+ key, m.name = f"path-{base.name}", rest
249
+ elif older:
250
+ raise CacheError(f"the archive's path {m.name} is not in this project or a "
251
+ "path dependency of it; build from source")
252
+ elif head == "project" and rest:
238
253
  key, base, m.name = "project", project, rest
239
254
  elif head == "path" and rest.partition("/")[0] in deps:
240
255
  name, _, m.name = rest.partition("/")
@@ -428,7 +443,7 @@ def restore(project: Path, repo: str | None = None, local: Path | None = None,
428
443
  src.fetch(name, tmp / name)
429
444
  man = json.loads((tmp / name).read_text())
430
445
  if not man.get("layout") and not name.startswith(_prefix(project.name)):
431
- continue # an archive in the older layout unpacks only into a folder of its name
446
+ continue # an archive in the older layout is named after the checkout's folder
432
447
  wanted = [p["name"] for p in man["parts"]] if man.get("parts") else [f"{stem}.tar.zst"]
433
448
  if any(w not in src.assets for w in wanted):
434
449
  continue
@@ -481,10 +496,7 @@ def restore(project: Path, repo: str | None = None, local: Path | None = None,
481
496
  if got != man["sha256"]:
482
497
  raise CacheError(f"sha256 {got} is not the manifest's {man['sha256']}; "
483
498
  "build from source")
484
- if man.get("layout") == 2:
485
- _unpack(arc, project)
486
- else:
487
- _sh(["tar", "-C", str(project.parent), "-I", "zstd -T0", "-xf", str(arc)])
499
+ _unpack(arc, project, older=man.get("layout") != 2)
488
500
  log(f"restored {stem} ("
489
501
  + ("the same Lean inputs" if key[0] == 0 else f"{dist} commits behind HEAD")
490
502
  + f", sha256 {got[:16]}…) "
@@ -352,6 +352,7 @@ def _check_json(a):
352
352
  "status": c["status"], "status_code": c["status_code"], "class": c["class"],
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
+ "correspondence": c.get("correspondence"),
355
356
  "suggestions": [{"name": x["name"], "why": x.get("why")}
356
357
  for x in c.get("suggestions", [])]}
357
358
  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"),
@@ -36,8 +36,16 @@ structure Query where
36
36
  accept_prefix : String
37
37
  allowed_axioms : Array String
38
38
  requests : Array Request
39
+ /-- Each request's heartbeat budget, in thousands as `maxHeartbeats`; Lean's default if none. -/
40
+ max_heartbeats : Option Nat := none
39
41
  deriving FromJson, Inhabited
40
42
 
43
+ /-- Runs `x` with a heartbeat budget of `n` thousand heartbeats when one is given. -/
44
+ def withBudget (n? : Option Nat) (x : MetaM α) : MetaM α :=
45
+ match n? with
46
+ | some n => withTheReader Core.Context (fun c => { c with maxHeartbeats := n * 1000 }) x
47
+ | none => x
48
+
41
49
  /-! ## Environment helpers -/
42
50
 
43
51
  def moduleOf (env : Environment) (n : Name) : Name :=
@@ -376,7 +384,8 @@ def emitClosure (h : IO.FS.Handle) (roots : Array Name) (root : Name) (used : Ar
376
384
  unless inScope roots env n do continue
377
385
  let some c := env.find? n | continue
378
386
  members := members.push n.toString
379
- let (j, more) ← declRecord roots n c
387
+ -- Each member has its own heartbeat budget: a closure can hold thousands.
388
+ let (j, more) ← withCurrHeartbeats (declRecord roots n c)
380
389
  unless (← seen.get).contains n do
381
390
  seen.modify (·.insert n)
382
391
  emit h j
@@ -479,7 +488,10 @@ def indexRecords (h : IO.FS.Handle) (roots : Array Name) : MetaM Unit := do
479
488
  | .axiomInfo _ => some "axiom" | _ => none
480
489
  let some k := kind | continue
481
490
  if k == "def" && (← isInstance n) then continue
482
- let ty ← try pure (toString (← ppExpr ci.type)) catch _ => pure "(not printable)"
491
+ -- Each declaration printed has its own heartbeat budget: a project can hold tens of
492
+ -- thousands, more than one budget prints.
493
+ let ty ← withCurrHeartbeats do
494
+ try pure (toString (← ppExpr ci.type)) catch _ => pure "(not printable)"
483
495
  let doc ← findDocString? env n
484
496
  emit h (Json.mkObj [("record", "index"), ("name", n.toString), ("kind", k),
485
497
  ("module", mod.toString), ("range", ← rangeJson n),
@@ -624,8 +636,10 @@ open ProofLensProbe in
624
636
  ("githash", Lean.githash)])
625
637
  layering h q
626
638
  for r in q.requests do
627
- -- Each request has its own heartbeat budget.
628
- try withCurrHeartbeats <|
639
+ -- Each request has its own heartbeat budget. A request that fails, including by running
640
+ -- out of heartbeats (a runtime exception, which `try` does not catch), emits an error
641
+ -- record for itself and the probe goes on.
642
+ tryCatchRuntimeEx (withBudget q.max_heartbeats <| withCurrHeartbeats do
629
643
  match r.kind with
630
644
  | "spec" => specRecord h q roots seen r.name.toName
631
645
  | "accept" | "refute" => proofRecord h q r.kind r.name.toName (r.spec.map (·.toName))
@@ -645,9 +659,9 @@ open ProofLensProbe in
645
659
  else
646
660
  emit h (Json.mkObj [("record", "defclosure"), ("name", r.name),
647
661
  ("errors", toJson #["missing"])])
648
- | _ => pure ()
649
- catch err =>
650
- emit h (Json.mkObj [("record", if r.kind == "decl" then "defclosure" else r.kind),
651
- ("name", r.name), ("spec", match r.spec with | some x => Json.str x | none => Json.null),
652
- ("errors", toJson #[s!"probe-error: {← err.toMessageData.toString}"])])
662
+ | _ => pure ())
663
+ (fun err => do
664
+ emit h (Json.mkObj [("record", if r.kind == "decl" then "defclosure" else r.kind),
665
+ ("name", r.name), ("spec", match r.spec with | some x => Json.str x | none => Json.null),
666
+ ("errors", toJson #[s!"probe-error: {← err.toMessageData.toString}"])]))
653
667
  emit h (Json.mkObj [("record", "done")])
@@ -105,7 +105,10 @@ def public_claim(chunk: dict, pub) -> dict:
105
105
  out["location"] = chunk.get("location") if pub.source_text else None
106
106
  out["statement_tex"] = chunk.get("statement_tex") if pub.source_text else None
107
107
  a = chunk.get("anchor")
108
- out["anchor"] = _pick(a, ["id", "env", "number", "label", "begin", "end", "kind", "source_sha256"] +
108
+ # `head` (the baseline of the first line) is what the index's top is computed from, so the
109
+ # published claim must keep it for its location to agree with the index.
110
+ out["anchor"] = _pick(a, ["id", "env", "number", "label", "begin", "head", "end", "kind",
111
+ "source_sha256"] +
109
112
  (["file", "line"] if pub.source_text else [])) if a else None
110
113
  clauses = []
111
114
  for cl in chunk["clauses"]:
@@ -279,7 +282,8 @@ def export(cfg: Config, out: Path, release: str | None = None,
279
282
  for p in index["papers"]:
280
283
  pub = pubs[p["id"]]
281
284
  q = _pick(p, ["id", "title", "mode", "revision", "pdf_sha256", "page_size", "note",
282
- "published", "published_note", "source_commit"])
285
+ "published", "published_note", "source_commit",
286
+ "paper_version"])
283
287
  q["root"] = p.get("root") if pub.source_text else None
284
288
  q["pdf"] = p["pdf"] if pub.pdf else None
285
289
  pidx["papers"].append(q)
@@ -64,9 +64,10 @@ def statements(fy: dict) -> list[dict]:
64
64
  continue
65
65
  rest = src[m.end():]
66
66
  module = r.get("module") if isinstance(r.get("module"), str) else None
67
+ note = r.get("note").strip() if isinstance(r.get("note"), str) else ""
67
68
  out.append({"source": src, "kind": m.group(1).lower(), "number": m.group(2),
68
69
  "hint": _tokens(rest), "lean": names, "status": r.get("status"),
69
- "module": module})
70
+ "module": module, "note": note})
70
71
  return out
71
72
 
72
73
 
@@ -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
@@ -74,7 +80,8 @@ def build(cfg: Config, release: str | None = None, skip_lean: bool = False,
74
80
  or (r["kind"] in ("accept", "refute", "witness") and
75
81
  f"{r['kind']}:{r['name']}:{r.get('spec') or ''}" not in lean["proofs"])]
76
82
  if missing:
77
- raise RuntimeError(f"the stored probe does not cover {missing[:5]}; run the probe")
83
+ raise StaleEvidence(f"the stored Lean results do not cover {', '.join(missing[:5])}"
84
+ + (", …" if len(missing) > 5 else "") + "; " + RUN_PROBE)
78
85
  lean_digest = lean_inputs_digest(cfg)
79
86
  log("probe: reusing the stored probe records (no Lean input changed)")
80
87
  elif not skip_lean:
@@ -93,7 +100,8 @@ def build(cfg: Config, release: str | None = None, skip_lean: bool = False,
93
100
  # A paper read from a folder: the commit of the repository that folder is in.
94
101
  **({"source_commit": git_head(cfg.root / paper.source["path"])}
95
102
  if snap.kind == "path" and git_head(cfg.root / paper.source["path"]) else {}),
96
- **({"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 {})}
97
105
  t = None
98
106
  if snap.mode == "latex" and not skip_tex:
99
107
  envs = paper.envs or sorted(set(DEFAULT_ENVS) | set(theorem_envs(snap.dir)))
@@ -214,16 +222,17 @@ def reusable_lean(cfg: Config) -> dict:
214
222
  content of every Lean input (here and in every path dependency) and the same probe."""
215
223
  rels = ev.releases(cfg.out_dir)
216
224
  if not rels:
217
- raise RuntimeError("no stored evidence to reuse")
225
+ raise StaleEvidence("no stored Lean results to reuse; " + RUN_PROBE)
218
226
  body = ev.load(cfg.out_dir, read_json(rels[-1])["evidence"])["body"]
219
227
  lean, prov = body.get("lean"), body["provenance"]
220
228
  if not lean or not prov.get("lean_inputs_sha256") or not prov.get("probe_sha256"):
221
- raise RuntimeError("the stored evidence does not record its Lean inputs; run the probe")
229
+ raise StaleEvidence("the stored Lean results do not record their Lean inputs; "
230
+ + RUN_PROBE)
222
231
  if prov["lean_inputs_sha256"] != lean_inputs_digest(cfg):
223
- raise RuntimeError("Lean inputs changed since the stored probe (here or in a path "
224
- "dependency); run the probe")
232
+ raise StaleEvidence("Lean inputs changed since the stored Lean results (here or in a "
233
+ "path dependency); " + RUN_PROBE)
225
234
  if prov["probe_sha256"] != probe_digest():
226
- raise RuntimeError("the probe changed since the stored records; run the probe")
235
+ raise StaleEvidence("the Lean probe changed since the stored Lean results; " + RUN_PROBE)
227
236
  return lean
228
237
 
229
238
 
@@ -242,6 +251,7 @@ class StaleEvidence(RuntimeError):
242
251
 
243
252
 
244
253
  NO_RELEASE = "no release: run `machinechecked build` first"
254
+ RUN_PROBE = "run `machinechecked build` without `--reuse-lean`"
245
255
 
246
256
 
247
257
  def inventory_changed(cfg: Config, body: dict) -> str | None:
@@ -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'}