machinechecked 0.1.9__tar.gz → 0.1.11__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 (71) hide show
  1. {machinechecked-0.1.9 → machinechecked-0.1.11}/PKG-INFO +1 -1
  2. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/machinechecked.egg-info/PKG-INFO +1 -1
  3. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/machinechecked.egg-info/SOURCES.txt +2 -2
  4. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/assess.py +29 -2
  5. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/cli.py +10 -2
  6. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/config.py +4 -1
  7. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/data/ProofLensProbe.lean +25 -7
  8. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/manifest.py +8 -2
  9. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/paper.py +56 -10
  10. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/pipeline.py +8 -1
  11. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/server.py +11 -6
  12. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/symbols.py +64 -10
  13. machinechecked-0.1.9/engine/python/prooflens/web/assets/index-CF9Ifrju.css → machinechecked-0.1.11/engine/python/prooflens/web/assets/index-B7NQsymB.css +1 -1
  14. machinechecked-0.1.11/engine/python/prooflens/web/assets/index-BtJVnEBn.js +286 -0
  15. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/index.html +2 -2
  16. {machinechecked-0.1.9 → machinechecked-0.1.11}/pyproject.toml +1 -1
  17. machinechecked-0.1.9/engine/python/prooflens/web/assets/index-CsVq8qV_.js +0 -286
  18. {machinechecked-0.1.9 → machinechecked-0.1.11}/LICENSE +0 -0
  19. {machinechecked-0.1.9 → machinechecked-0.1.11}/README.md +0 -0
  20. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/machinechecked.egg-info/dependency_links.txt +0 -0
  21. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/machinechecked.egg-info/entry_points.txt +0 -0
  22. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/machinechecked.egg-info/requires.txt +0 -0
  23. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/machinechecked.egg-info/top_level.txt +0 -0
  24. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/__init__.py +0 -0
  25. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/__main__.py +0 -0
  26. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/audit.py +0 -0
  27. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/bundle.py +0 -0
  28. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/cache.py +0 -0
  29. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/data/prooflens.sty +0 -0
  30. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/data/schema/evidence.schema.json +0 -0
  31. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/data/schema/inventory.schema.json +0 -0
  32. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/data/schema/manifest-claim.schema.json +0 -0
  33. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/data/schema/manifest-index.schema.json +0 -0
  34. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/deps.py +0 -0
  35. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/evidence.py +0 -0
  36. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/export.py +0 -0
  37. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/formalization.py +0 -0
  38. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/pdfclaims.py +0 -0
  39. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/probe.py +0 -0
  40. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/promote.py +0 -0
  41. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/replay.py +0 -0
  42. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/review_cli.py +0 -0
  43. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/review_queue.py +0 -0
  44. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/reviews.py +0 -0
  45. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/schemas.py +0 -0
  46. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/simple.py +0 -0
  47. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/sources.py +0 -0
  48. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/tex.py +0 -0
  49. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/util.py +0 -0
  50. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_AMS-Regular-BQhdFMY1.woff2 +0 -0
  51. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Caligraphic-Bold-Dq_IR9rO.woff2 +0 -0
  52. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Caligraphic-Regular-Di6jR-x-.woff2 +0 -0
  53. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Fraktur-Bold-CL6g_b3V.woff2 +0 -0
  54. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Fraktur-Regular-CTYiF6lA.woff2 +0 -0
  55. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Main-Bold-Cx986IdX.woff2 +0 -0
  56. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Main-BoldItalic-DxDJ3AOS.woff2 +0 -0
  57. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Main-Italic-NWA7e6Wa.woff2 +0 -0
  58. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Main-Regular-B22Nviop.woff2 +0 -0
  59. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Math-BoldItalic-CZnvNsCZ.woff2 +0 -0
  60. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Math-Italic-t53AETM-.woff2 +0 -0
  61. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_SansSerif-Bold-D1sUS0GD.woff2 +0 -0
  62. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_SansSerif-Italic-C3H0VqGB.woff2 +0 -0
  63. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_SansSerif-Regular-DDBCnlJ7.woff2 +0 -0
  64. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Script-Regular-D3wIWfF6.woff2 +0 -0
  65. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Size1-Regular-mCD8mA8B.woff2 +0 -0
  66. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Size2-Regular-Dy4dx90m.woff2 +0 -0
  67. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Size4-Regular-Dl5lxZxV.woff2 +0 -0
  68. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/KaTeX_Typewriter-Regular-CO6r4hn1.woff2 +0 -0
  69. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/assets/pdf.worker.min-yatZIOMy.mjs +0 -0
  70. {machinechecked-0.1.9 → machinechecked-0.1.11}/engine/python/prooflens/web/theme.css +0 -0
  71. {machinechecked-0.1.9 → machinechecked-0.1.11}/setup.cfg +0 -0
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: machinechecked
3
- Version: 0.1.9
3
+ Version: 0.1.11
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: 0.1.9
3
+ Version: 0.1.11
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
@@ -63,6 +63,6 @@ engine/python/prooflens/web/assets/KaTeX_Size1-Regular-mCD8mA8B.woff2
63
63
  engine/python/prooflens/web/assets/KaTeX_Size2-Regular-Dy4dx90m.woff2
64
64
  engine/python/prooflens/web/assets/KaTeX_Size4-Regular-Dl5lxZxV.woff2
65
65
  engine/python/prooflens/web/assets/KaTeX_Typewriter-Regular-CO6r4hn1.woff2
66
- engine/python/prooflens/web/assets/index-CF9Ifrju.css
67
- engine/python/prooflens/web/assets/index-CsVq8qV_.js
66
+ engine/python/prooflens/web/assets/index-B7NQsymB.css
67
+ engine/python/prooflens/web/assets/index-BtJVnEBn.js
68
68
  engine/python/prooflens/web/assets/pdf.worker.min-yatZIOMy.mjs
@@ -442,7 +442,10 @@ class Assessor:
442
442
  consistency = [f["message"] for f in mine]
443
443
  reasons = []
444
444
  t = (self.tex or {}).get(pid, {})
445
- if t.get("mode") != "latex":
445
+ source = t.get("mode") or "none"
446
+ if source == "pdf":
447
+ reasons += self._pdf_source(pid, t)
448
+ elif source != "latex":
446
449
  reasons.append("sources: manuscript source not compiled, source hashes unchecked")
447
450
  if self.lean is None:
448
451
  reasons.append("lean: the Lean project was not checked")
@@ -472,11 +475,35 @@ class Assessor:
472
475
  if not comparator:
473
476
  reasons.append("comparator: no successful kernel replay covering every proof "
474
477
  "of the paper")
475
- out[pid] = {"consistency": {"pass": not consistency, "failures": consistency},
478
+ out[pid] = {"source": source,
479
+ **({"source_note": "the source is a PDF: its hash is pinned and checked, "
480
+ "and claim positions are approximate"} if source == "pdf" else {}),
481
+ "consistency": {"pass": not consistency, "failures": consistency},
476
482
  "release": {"pass": not reasons, "reasons": reasons},
477
483
  "comparator": comparator}
478
484
  return out
479
485
 
486
+ def _pdf_source(self, pid: str, t: dict) -> list[str]:
487
+ """Why a paper whose source is a PDF does not meet the gate's source condition: the PDF
488
+ is pinned by the configuration (`pdf_sha256`) and matches it, and every claim in scope
489
+ is found in it (by its heading, or placed by hand in `anchors.json`)."""
490
+ out = []
491
+ pinned, actual = t.get("pdf_pinned"), t.get("pdf_sha256")
492
+ if not pinned:
493
+ out.append("sources: the paper is a PDF and its hash is not pinned (pdf_sha256 in "
494
+ "machinechecked.toml), so it is not checked")
495
+ elif pinned != actual:
496
+ out.append(f"sources: the PDF differs from the one pinned (sha256 {str(actual)[:12]}, "
497
+ f"pinned {pinned[:12]})")
498
+ manual = t.get("manual_anchors") or {}
499
+ for c in self.inv["claims"]:
500
+ if c["paper"] != pid or c["scope"] in OUT_OF_SCOPE:
501
+ continue
502
+ if c["id"] not in manual and match_anchor(c, t.get("anchors", [])) is None:
503
+ out.append(f"sources: {c['id']} ({c.get('number') or 'unnumbered'}) is not found "
504
+ "in the PDF; place it by hand in anchors.json")
505
+ return out
506
+
480
507
  # -- inventory against the compiled paper ---------------------------------------------
481
508
 
482
509
  def inventory_checks(self):
@@ -228,7 +228,12 @@ def cmd_view(a):
228
228
  res = build(cfg, web_dist=_web_dist(a.web), reuse_lean=True)
229
229
  except RuntimeError:
230
230
  res = build(cfg, web_dist=_web_dist(a.web))
231
- serve(res["site"], a.port, None, not a.no_browser)
231
+
232
+ def on_symbol(req):
233
+ from .symbols import from_reader
234
+ from_reader(cfg, req, _web_dist(a.web))
235
+
236
+ serve(res["site"], a.port, None, not a.no_browser, on_symbol=on_symbol)
232
237
  return
233
238
  from .server import serve
234
239
  from .simple import save_link
@@ -531,7 +536,8 @@ def cmd_symbol(a):
531
536
  cfg = cfgmod.load(_config_arg(a))
532
537
  try:
533
538
  if a.action == "add":
534
- ln = sy.add(cfg, a.claim, a.tex, a.decl, a.reviewer, var=a.var, type_of=a.type_of)
539
+ ln = sy.add(cfg, a.claim, a.tex, a.decl, a.reviewer, var=a.var, type_of=a.type_of,
540
+ term=a.term)
535
541
  print(f"linked {a.tex} in {a.claim} to {ln['lean']} (by {ln['by']}); "
536
542
  "rebuild the site to show it")
537
543
  elif a.action == "remove":
@@ -802,6 +808,8 @@ def main(argv=None):
802
808
  p_.add_argument("decl", help="the claim's Lean statement")
803
809
  p_.add_argument("--var", help="the variable it is (a binder's name or index)")
804
810
  p_.add_argument("--type-of", dest="type_of", help="the binder whose type it is")
811
+ p_.add_argument("--term", help="the sub-term it is, as WHERE#K: b:<binder>, c (the "
812
+ "conclusion) or l:<part>, and the sub-term's number")
805
813
  p_.add_argument("--as", dest="reviewer",
806
814
  default=os.environ.get("MACHINECHECKED_REVIEWER"),
807
815
  help="human:NAME or llm:MODEL (default: $MACHINECHECKED_REVIEWER)")
@@ -38,6 +38,9 @@ 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
+ # A paper whose source is a PDF: the PDF's SHA-256, pinned by the configuration. The release
42
+ # gate accepts a PDF as source evidence only when it is pinned and matches (EVIDENCE_POLICY).
43
+ pdf_sha256: str | None = None
41
44
 
42
45
 
43
46
  @dataclass
@@ -117,7 +120,7 @@ def load(path: str | Path) -> Config:
117
120
  raise ConfigError(f"paper {p.get('id')}: source needs one of path, git, arxiv")
118
121
  papers.append(Paper(
119
122
  id=_req(p, "id", "paper"), title=p.get("title", p["id"]), source=src,
120
- root=p.get("root"), pdf=p.get("pdf"), commit=src.get("rev"), envs=p.get("envs"),
123
+ root=p.get("root"), pdf=p.get("pdf"), pdf_sha256=p.get("pdf_sha256"), commit=src.get("rev"), envs=p.get("envs"),
121
124
  publish=Publish(**p.get("publish", {})), published=p.get("published"),
122
125
  published_note=p.get("note")))
123
126
  return Config(
@@ -167,15 +167,18 @@ def constPrint (ci : ConstantInfo) : MetaM (String × String × Array Name) := d
167
167
 
168
168
  /-- Flatten tagged code into segments; a segment carries the constant it names (`ref`), or the
169
169
  index of the statement's binder it refers to (`var`, a position in `vars`), when the innermost
170
- subexpression around it is one. -/
170
+ subexpression around it is one. Each tagged subexpression is also recorded as the range of
171
+ segments it covers (`terms`, in printing order, an outer one before those inside it), so that the
172
+ reader can select a whole sub-term. -/
171
173
  partial def segments (vars : Array FVarId) (t : Widget.CodeWithInfos) (ref : Option (String × Json))
172
- (acc : Array Json) : Array Json :=
174
+ (acc : Array Json) (terms : Array (Nat × Nat)) : Array Json × Array (Nat × Nat) :=
173
175
  match t with
174
176
  | Widget.TaggedText.text s =>
175
177
  match ref with
176
- | some (k, v) => acc.push (Json.mkObj [("t", s), (k, v)])
177
- | none => acc.push (Json.mkObj [("t", s)])
178
- | Widget.TaggedText.append xs => xs.foldl (fun a x => segments vars x ref a) acc
178
+ | some (k, v) => (acc.push (Json.mkObj [("t", s), (k, v)]), terms)
179
+ | none => (acc.push (Json.mkObj [("t", s)]), terms)
180
+ | Widget.TaggedText.append xs =>
181
+ xs.foldl (fun (a, ts) x => segments vars x ref a ts) (acc, terms)
179
182
  | Widget.TaggedText.tag info sub =>
180
183
  let r := match info.info.val.info with
181
184
  | .ofTermInfo ti => match ti.expr with
@@ -183,7 +186,21 @@ partial def segments (vars : Array FVarId) (t : Widget.CodeWithInfos) (ref : Opt
183
186
  | .fvar id => (vars.findIdx? (· == id)).map fun i => ("var", toJson i)
184
187
  | _ => none
185
188
  | _ => none
186
- segments vars sub r acc
189
+ let start := acc.size
190
+ let slot := terms.size
191
+ let (acc', terms') := segments vars sub r acc (terms.push (start, start))
192
+ (acc', terms'.set! slot (start, acc'.size))
193
+
194
+ /-- The sub-terms worth selecting: those covering two segments or more, each range once (the
195
+ outermost tag of a range keeps it), and none for a very large term. -/
196
+ def termsJson (terms : Array (Nat × Nat)) : Json := Id.run do
197
+ let mut out : Array Json := #[]
198
+ let mut seen : Array (Nat × Nat) := #[]
199
+ for (a, b) in terms do
200
+ if b - a ≥ 2 && !seen.contains (a, b) then
201
+ seen := seen.push (a, b)
202
+ out := out.push (Json.mkObj [("from", toJson a), ("to", toJson b)])
203
+ return Json.arr (if out.size > 3000 then #[] else out)
187
204
 
188
205
  /-- The fields of a pretty-printed expression with navigable constants and binders; plain text
189
206
  if tagging fails, and a marker if printing fails, so that one large term cannot stop the probe. -/
@@ -193,7 +210,8 @@ def ppFields (e : Expr) (vars : Array FVarId := #[]) : MetaM (List (String × Js
193
210
  -- term applies is a constant the reader can name; the text is the same.
194
211
  let code ← withOptions (fun o => o.setBool `pp.tagAppFns true) (Widget.ppExprTagged e)
195
212
  let text := toString (← ppExpr e)
196
- return [("text", Json.str text), ("segments", Json.arr (segments vars code none #[]))]
213
+ let (segs, terms) := segments vars code none #[] #[]
214
+ return [("text", Json.str text), ("segments", Json.arr segs), ("terms", termsJson terms)]
197
215
  catch _ =>
198
216
  try
199
217
  return [("text", Json.str (toString (← ppExpr e)))]
@@ -175,9 +175,15 @@ def build(evidence: dict, release: str, display: dict | None = None
175
175
  for c in inv:
176
176
  a = assessed[c["id"]]
177
177
  t = tex.get(c["paper"], {})
178
- anchor = match_anchor(c, t.get("anchors", [])) if t.get("mode") == "latex" else None
178
+ anchor = match_anchor(c, t.get("anchors", [])) if t.get("mode") in ("latex", "pdf") \
179
+ else None
179
180
  src_ok = None
180
- if anchor:
181
+ if anchor and t.get("mode") == "pdf":
182
+ # Found by its heading in the PDF: where it ends is estimated.
183
+ anchor = {**anchor, "kind": "approximate"}
184
+ elif anchor:
185
+ # (The inventory's source hash is of the LaTeX statement; a PDF's extracted text
186
+ # cannot be compared with it.)
181
187
  anchor = {**anchor, "kind": "exact"}
182
188
  if c.get("source_sha256") and anchor.get("source_sha256"):
183
189
  src_ok = anchor["source_sha256"] == c["source_sha256"]
@@ -4,8 +4,11 @@ states it formalises the paper, the declarations its `formalization.yaml` maps t
4
4
  It is written in the reader's standard data format (`data/index.json`, `data/claims/`), so a
5
5
  paper page is read by the same code as a view or an audit. Nothing here runs Lean, so a mapped
6
6
  statement reads "linked: Lean not checked here (no audit published)": the link says which
7
- declaration the project names, not what Lean establishes. Each project's links and statuses are
8
- kept apart (`by_project`); the reader shows one project at a time and never combines them.
7
+ declaration the project names, not what Lean establishes. When the project has a published
8
+ audit that lists a claim of the same printed number, the statement shows that claim's status as
9
+ the audit records it, credited to the audit's author and linked to it (`apply_audits`): the
10
+ audit's words, not the site's. Each project's links and statuses are kept apart (`by_project`);
11
+ the reader shows one project at a time and never combines them.
9
12
 
10
13
  No statement's text is written: only its number, heading title and place on the page.
11
14
  """
@@ -21,6 +24,7 @@ from .util import content_id
21
24
 
22
25
  SCHEMA = "prooflens/manifest@0"
23
26
  NOT_CHECKED = "linked: Lean not checked here (no audit published)"
27
+ NOT_IN_AUDIT = "linked: Lean not checked here (the published audit does not list it)"
24
28
  NOT_LINKED = "not linked"
25
29
  NAME = re.compile(r"[\w.'«»]+")
26
30
 
@@ -47,8 +51,44 @@ def paper_claims(anchors: list[dict], title: str,
47
51
  return out
48
52
 
49
53
 
50
- def _state(links: list[dict]) -> tuple[str, str]:
51
- return (NOT_CHECKED, "declared") if links else (NOT_LINKED, "unspecified")
54
+ def apply_audits(claims: list[dict], audits: dict[str, dict]) -> list[dict]:
55
+ """Each statement (from `paper_claims`) with, for every project whose published audit lists
56
+ exactly one claim of the same printed number, that claim as the audit records it (`audit`:
57
+ status, class, claim id, link, author). When the project's formalization.yaml declares no
58
+ link for the statement, the audit's Lean names are its links. `audits` maps a repository to
59
+ `{"author", "url", "commit", "claims": [{"id", "number", "status", "class", "lean",
60
+ "link"}]}`, read from the audit's own data; the claims are not changed otherwise."""
61
+ for c in claims:
62
+ c["audit"] = {}
63
+ for repo, au in audits.items():
64
+ same = [x for x in au.get("claims", []) if x.get("number") == c["number"]]
65
+ if len(same) != 1:
66
+ continue
67
+ x = same[0]
68
+ c["audit"][repo] = {"status": x["status"], "class": x["class"], "claim": x["id"],
69
+ "link": x.get("link"), "author": au.get("author"),
70
+ "audit": au.get("url")}
71
+ if not any(l["repository"] == repo for l in c["links"]):
72
+ c["links"] = c["links"] + [
73
+ {"name": n, "repository": repo, "commit": au.get("commit"),
74
+ "source": "audit"} for n in x.get("lean", []) if NAME.fullmatch(n)]
75
+ return claims
76
+
77
+
78
+ def _state(links: list[dict], audit: dict | None = None,
79
+ audited: bool = False) -> tuple[str, str]:
80
+ """A statement's status for one project: the audit's own, when its published audit lists
81
+ the statement; otherwise what the page can say without Lean."""
82
+ if audit:
83
+ return audit["status"], audit["class"]
84
+ if not links:
85
+ return NOT_LINKED, "unspecified"
86
+ return (NOT_IN_AUDIT if audited else NOT_CHECKED), "declared"
87
+
88
+
89
+ def _link_state(link: dict, audit: dict | None, audited: bool) -> dict:
90
+ status, cls = _state([link], audit, audited)
91
+ return {"status": status, "class": cls}
52
92
 
53
93
 
54
94
  def write_paper_site(site: Path, paper: dict, claims: list[dict],
@@ -60,17 +100,22 @@ def write_paper_site(site: Path, paper: dict, claims: list[dict],
60
100
  `paper_claims`. Replaces whatever is at `site`."""
61
101
  from .pipeline import write_preview
62
102
  projects = [f["repository"] for f in paper.get("formalisations", [])]
103
+ audited = {f["repository"] for f in paper.get("formalisations", []) if f.get("audits")}
63
104
  index_claims, chunks = [], {}
64
105
  for c in claims:
65
106
  cid = f"P:{c['key']}"
66
107
  fname = mf.claim_file(cid)
67
108
  anchor = {**c["anchor"], "kind": "approximate"} if c.get("anchor") else None
68
109
  by_project = {}
110
+ audits = c.get("audit") or {}
69
111
  for r in projects:
70
- status, cls = _state([l for l in c["links"] if l["repository"] == r])
71
- by_project[r] = {"status": status, "class": cls}
72
- status, cls = _state([l for l in c["links"]
73
- if not projects or l["repository"] == projects[0]])
112
+ status, cls = _state([l for l in c["links"] if l["repository"] == r], audits.get(r),
113
+ r in audited)
114
+ by_project[r] = {"status": status, "class": cls,
115
+ **({"audit": audits[r]} if r in audits else {})}
116
+ first = projects[0] if projects else None
117
+ status, cls = _state([l for l in c["links"] if not projects or l["repository"] == first],
118
+ audits.get(first), first in audited)
74
119
  common = {"id": cid, "paper": "P", "number": c["number"], "title": c.get("title"),
75
120
  "heading": c.get("title") or None, "kind": c["kind"], "scope": "formal",
76
121
  "status": status, "class": cls, "by_project": by_project}
@@ -81,8 +126,9 @@ def write_paper_site(site: Path, paper: dict, claims: list[dict],
81
126
  **common, "schema": SCHEMA, "mode": "simple", "release": "paper", "key": c["key"],
82
127
  "scope_note": "", "scope_reviewed": True, "anchor": anchor,
83
128
  "source_matches": None, "statement_tex": None, "blockers": [],
84
- "links": [{**l, "status": NOT_CHECKED, "class": "declared", "why": None,
85
- "statement": None, "path": None} for l in c["links"]],
129
+ "links": [{**l, **_link_state(l, audits.get(l["repository"]),
130
+ l["repository"] in audited),
131
+ "why": None, "statement": None, "path": None} for l in c["links"]],
86
132
  "suggestions": [], "clauses": [], "decls": {}, "findings": []}
87
133
  summary = {}
88
134
  for c in index_claims:
@@ -112,7 +112,14 @@ def build(cfg: Config, release: str | None = None, skip_lean: bool = False,
112
112
  prov["pdf_origin"] = "compiled"
113
113
  pdfs[paper.id] = t["pdf"]
114
114
  elif snap.pdf:
115
- tex[paper.id] = {"mode": "pdf", "anchors": [], "page_size": None, "inputs": []}
115
+ # The source is a PDF: its numbered statements are found by their bold headings
116
+ # (positions approximate). The PDF counts as source evidence for the release gate
117
+ # only when the configuration pins its hash and it matches.
118
+ from .pdfclaims import claims_from_pdf
119
+ anchors, size = claims_from_pdf(snap.dir / snap.pdf)
120
+ tex[paper.id] = {"mode": "pdf", "anchors": anchors, "page_size": size, "inputs": [],
121
+ "pdf_sha256": prov["committed_pdf_sha256"],
122
+ "pdf_pinned": paper.pdf_sha256}
116
123
  prov["mode"] = "pdf"
117
124
  prov["pdf_sha256"] = prov["committed_pdf_sha256"]
118
125
  prov["pdf_origin"] = "committed"
@@ -2,7 +2,9 @@
2
2
 
3
3
  It listens on 127.0.0.1 only. `GET /api/status` tells the reader it may edit; `POST /api/link`
4
4
  with `{"key", "lean": [names] | null, "reject": [names]}` records a link (or removes it, or
5
- rejects suggestions) in `prooflens.links.json` and rebuilds the view.
5
+ rejects suggestions) in `prooflens.links.json` and rebuilds the view. In a full assessment,
6
+ `POST /api/symbol` with `{"action": "add" | "remove", "claim", "tex", "decl", "var" | "type_of",
7
+ "by"}` records or removes a symbol link (docs/SYMBOL_LINKS.md) and rewrites the site.
6
8
  """
7
9
 
8
10
  from __future__ import annotations
@@ -14,8 +16,10 @@ import threading
14
16
  from pathlib import Path
15
17
 
16
18
 
17
- def serve(site: Path, port: int, on_link=None, open_browser: bool = True):
19
+ def serve(site: Path, port: int, on_link=None, open_browser: bool = True, on_symbol=None):
18
20
  lock = threading.Lock()
21
+ editable = on_link is not None or on_symbol is not None
22
+ routes = {"/api/link": on_link, "/api/symbol": on_symbol}
19
23
 
20
24
  class Handler(http.server.SimpleHTTPRequestHandler):
21
25
  def log_message(self, *a):
@@ -31,10 +35,10 @@ def serve(site: Path, port: int, on_link=None, open_browser: bool = True):
31
35
 
32
36
  def do_GET(self):
33
37
  if self.path == "/api/status":
34
- return self._json(200, {"editable": on_link is not None})
38
+ return self._json(200, {"editable": editable})
35
39
  # The data index says the page may edit (only this local server says so; a static
36
40
  # copy of the site never asks a server anything).
37
- if self.path.split("?")[0] == "/data/index.json" and on_link is not None:
41
+ if self.path.split("?")[0] == "/data/index.json" and editable:
38
42
  try:
39
43
  data = json.loads((site / "data" / "index.json").read_text(encoding="utf-8"))
40
44
  except (OSError, ValueError):
@@ -43,13 +47,14 @@ def serve(site: Path, port: int, on_link=None, open_browser: bool = True):
43
47
  return super().do_GET()
44
48
 
45
49
  def do_POST(self):
46
- if self.path != "/api/link" or on_link is None:
50
+ handle = routes.get(self.path)
51
+ if handle is None:
47
52
  return self._json(404, {"error": "not found"})
48
53
  try:
49
54
  n = int(self.headers.get("Content-Length", "0"))
50
55
  req = json.loads(self.rfile.read(n) or b"{}")
51
56
  with lock:
52
- on_link(req)
57
+ handle(req)
53
58
  return self._json(200, {"ok": True})
54
59
  except Exception as e: # the reader shows the message
55
60
  return self._json(400, {"error": str(e)})
@@ -36,16 +36,41 @@ def _in_math(tex: str, symbol: str) -> bool:
36
36
  return any(re.search(pat, s) for s in spans)
37
37
 
38
38
 
39
+ def _term_pp(spec: dict, where: str) -> dict | None:
40
+ """The pretty-printed term a sub-term's place names: `b:<i>` a binder's type, `c` the
41
+ conclusion, `l:<n>` the n-th part of a conclusion split by its logic."""
42
+ concl = spec.get("conclusion") or {}
43
+ if where == "c":
44
+ return concl
45
+ if where.startswith("b:") and where[2:].isdigit():
46
+ bs = spec.get("binders", [])
47
+ i = int(where[2:])
48
+ return bs[i].get("type") if i < len(bs) else None
49
+ if where.startswith("l:") and where[2:].isdigit():
50
+ leaves = []
51
+
52
+ def walk(n):
53
+ if "leaf" in n:
54
+ leaves.append(n["leaf"])
55
+ for a in n.get("args", []):
56
+ walk(a)
57
+ if concl.get("logic"):
58
+ walk(concl["logic"])
59
+ n = int(where[2:])
60
+ return leaves[n] if n < len(leaves) else None
61
+ return None
62
+
63
+
39
64
  def add(cfg: Config, claim: str, symbol: str, decl: str, by: str, var: str | None = None,
40
- type_of: str | None = None) -> dict:
65
+ type_of: str | None = None, term: str | None = None) -> dict:
41
66
  """Record a link from `symbol` in `claim`'s printed statement to the variable `var` (or the
42
67
  type of the binder `type_of`) of the Lean statement `decl`, by `by`."""
43
68
  from .pipeline import lean_inputs_digest
44
69
  from .promote import latest_record
45
70
  if not re.fullmatch(r"(human|llm):\S+", by or ""):
46
71
  raise SymbolError("name the link's author as human:NAME or llm:MODEL (--as)")
47
- if (var is None) == (type_of is None):
48
- raise SymbolError("give exactly one of --var NAME and --type-of NAME")
72
+ if sum(x is not None for x in (var, type_of, term)) != 1:
73
+ raise SymbolError("give exactly one of --var NAME, --type-of NAME and --term WHERE#K")
49
74
  record = latest_record(cfg)
50
75
  body = record["body"]
51
76
  if body["provenance"].get("lean_inputs_sha256") != lean_inputs_digest(cfg):
@@ -62,13 +87,21 @@ def add(cfg: Config, claim: str, symbol: str, decl: str, by: str, var: str | Non
62
87
  raise SymbolError(f"{symbol!r} does not occur in the mathematics of {claim}'s printed "
63
88
  "statement")
64
89
  spec = (body.get("lean") or {}).get("specs", {}).get(decl) or {}
65
- names = [b.get("name") for b in spec.get("binders", [])]
66
- want = var if var is not None else type_of
67
- idx = int(want) if want.isdigit() else (names.index(want) if want in names else None)
68
- if idx is None or not 0 <= idx < len(names):
69
- raise SymbolError(f"{decl} has no binder {want!r} (it has {', '.join(names)})")
70
- target = {"decl": decl, "var": idx} if var is not None else \
71
- {"decl": decl, "binder": idx, "path": "type"}
90
+ if term is not None:
91
+ where, _, k = term.partition("#")
92
+ pp = _term_pp(spec, where)
93
+ if pp is None or not k.isdigit() or int(k) >= len(pp.get("terms") or []):
94
+ raise SymbolError(f"{decl} has no sub-term {term!r} (WHERE#K: b:<binder>, c or "
95
+ "l:<part>, then the sub-term's number)")
96
+ target = {"decl": decl, "term": int(k), "in": where}
97
+ else:
98
+ names = [b.get("name") for b in spec.get("binders", [])]
99
+ want = var if var is not None else type_of
100
+ idx = int(want) if want.isdigit() else (names.index(want) if want in names else None)
101
+ if idx is None or not 0 <= idx < len(names):
102
+ raise SymbolError(f"{decl} has no binder {want!r} (it has {', '.join(names)})")
103
+ target = {"decl": decl, "var": idx} if var is not None else \
104
+ {"decl": decl, "binder": idx, "path": "type"}
72
105
  link = {"claim": claim, "symbol": {"tex": symbol, "occurrence": "all"}, "lean": target,
73
106
  "by": by, "pins": {"statement": spec.get("fingerprint"), "source": sha256_text(tex)}}
74
107
  data = read_json(cfg.symbols) if cfg.symbols.exists() else {"schema": SCHEMA, "links": []}
@@ -95,3 +128,24 @@ def listing(cfg: Config) -> str:
95
128
  return "no symbol links"
96
129
  return "\n".join(f"{ln['claim']} {ln['symbol']['tex']} -> {json.dumps(ln['lean'])} "
97
130
  f"(by {ln['by']})" for ln in read_json(cfg.symbols).get("links", []))
131
+
132
+
133
+ def from_reader(cfg: Config, req: dict, web_dist=None) -> None:
134
+ """A symbol link made or removed in the local reader (`POST /api/symbol`), with the checks
135
+ of `symbol add`; the site is then rewritten from the stored evidence (links are display
136
+ only, so nothing is probed or assessed again)."""
137
+ from .pipeline import rebuild_site
138
+
139
+ def text(v):
140
+ return None if v is None else str(v)
141
+ if not isinstance(req.get("claim"), str) or not isinstance(req.get("tex"), str):
142
+ raise SymbolError("a symbol request names a claim and a symbol")
143
+ if req.get("action") == "remove":
144
+ remove(cfg, req["claim"], req["tex"])
145
+ elif req.get("action") == "add":
146
+ add(cfg, req["claim"], req["tex"], str(req.get("decl") or ""), req.get("by"),
147
+ var=text(req.get("var")), type_of=text(req.get("type_of")),
148
+ term=text(req.get("term")))
149
+ else:
150
+ raise SymbolError(f"unknown action {req.get('action')!r}")
151
+ rebuild_site(cfg, web_dist, None)