machinechecked 0.1.3__tar.gz → 0.1.4__tar.gz

{machinechecked-0.1.3 → machinechecked-0.1.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: machinechecked
-Version: 0.1.3
+Version: 0.1.4
 Summary: Read a mathematical paper beside its Lean formalisation, and see claim by claim what Lean establishes.
 License: Apache-2.0
 Project-URL: Homepage, https://machinechecked.org

{machinechecked-0.1.3 → machinechecked-0.1.4}/engine/python/machinechecked.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: machinechecked
-Version: 0.1.3
+Version: 0.1.4
 Summary: Read a mathematical paper beside its Lean formalisation, and see claim by claim what Lean establishes.
 License: Apache-2.0
 Project-URL: Homepage, https://machinechecked.org

{machinechecked-0.1.3 → machinechecked-0.1.4}/engine/python/machinechecked.egg-info/SOURCES.txt

@@ -18,6 +18,7 @@ engine/python/prooflens/evidence.py
 engine/python/prooflens/export.py
 engine/python/prooflens/formalization.py
 engine/python/prooflens/manifest.py
+engine/python/prooflens/paper.py
 engine/python/prooflens/pdfclaims.py
 engine/python/prooflens/pipeline.py
 engine/python/prooflens/probe.py
@@ -56,6 +57,6 @@ engine/python/prooflens/web/assets/KaTeX_Size1-Regular-mCD8mA8B.woff2
 engine/python/prooflens/web/assets/KaTeX_Size2-Regular-Dy4dx90m.woff2
 engine/python/prooflens/web/assets/KaTeX_Size4-Regular-Dl5lxZxV.woff2
 engine/python/prooflens/web/assets/KaTeX_Typewriter-Regular-CO6r4hn1.woff2
-engine/python/prooflens/web/assets/index-CEQVnsCC.js
+engine/python/prooflens/web/assets/index-Crg6e6Ow.js
 engine/python/prooflens/web/assets/index-T9UdK0vf.css
 engine/python/prooflens/web/assets/pdf.worker.min-yatZIOMy.mjs

{machinechecked-0.1.3 → machinechecked-0.1.4}/engine/python/prooflens/data/schema/manifest-index.schema.json

@@ -103,7 +103,8 @@
               "open",
               "malformed",
               "out-of-scope",
-              "unspecified"
+              "unspecified",
+              "declared"
             ]
           },
           "anchor": {

{machinechecked-0.1.3 → machinechecked-0.1.4}/engine/python/prooflens/manifest.py

@@ -45,6 +45,18 @@ def top_of(anchor: dict) -> tuple[int, float]:
     return anchor["begin"][0], anchor["begin"][2]
 
 
+def index_anchor(anchor: dict | None, kind: str) -> dict | None:
+    """A located claim's place as the index lists it (for the contents and the margin): its
+    page and top, where it ends on that page, and the page it ends on."""
+    if not anchor or not anchor.get("begin"):
+        return None
+    page, y0 = top_of({**anchor, "kind": kind})
+    end = anchor.get("end")
+    return {"kind": kind, "page": page, "y0": y0,
+            "y1": end[2] if end and end[0] == anchor["begin"][0] else None,
+            "page_end": end[0] if end else None}
+
+
 def ladder(cl: dict, anchor: dict | None, mode: str | None, src_ok: bool | None) -> list[dict]:
     """The rungs of the evidence ladder of one clause, each with its own state."""
     ch = cl.get("checks", {})

machinechecked-0.1.4/engine/python/prooflens/paper.py

@@ -0,0 +1,103 @@
+"""A paper's page: the paper's PDF, the statements found in it, and for each Lean project that
+states it formalises the paper, the declarations its `formalization.yaml` maps them to.
+
+It is written in the reader's standard data format (`data/index.json`, `data/claims/`), so a
+paper page is read by the same code as a view or an audit. Nothing here runs Lean, so a mapped
+statement reads "linked: Lean not checked here (no audit published)": the link says which
+declaration the project names, not what Lean establishes. Each project's links and statuses are
+kept apart (`by_project`); the reader shows one project at a time and never combines them.
+
+No statement's text is written: only its number, heading title and place on the page.
+"""
+
+from __future__ import annotations
+
+import re
+from pathlib import Path
+
+from . import formalization as fz
+from . import manifest as mf
+from .util import content_id
+
+SCHEMA = "prooflens/manifest@0"
+NOT_CHECKED = "linked: Lean not checked here (no audit published)"
+NOT_LINKED = "not linked"
+NAME = re.compile(r"[\w.'«»]+")
+
+
+def paper_claims(anchors: list[dict], title: str,
+                 alignments: list[tuple[dict, dict | None]]) -> list[dict]:
+    """The statements found in the paper (anchors from `pdfclaims.claims_from_pdf`), each with
+    the links the projects' `formalization.yaml` files declare for it. `alignments` pairs each
+    project (`repository`, `commit`) with its parsed `formalization.yaml`, or None when it has
+    none or it could not be read."""
+    out = []
+    for a in anchors:
+        number = f"{a['env'].capitalize()} {a['number']}"
+        links = []
+        for f, fy in alignments:
+            for row in fz.links_for(fy, a["env"], number, title) if fy else []:
+                links += [{"name": n, "repository": f["repository"], "commit": f["commit"],
+                           "source": "formalization.yaml"}
+                          for n in row["lean"] if NAME.fullmatch(n)]
+        out.append({"key": f"{a['env']}-{a['number']}", "number": number, "kind": a["env"],
+                    "title": a.get("title"),
+                    "anchor": {k: a[k] for k in ("begin", "end", "head") if a.get(k)},
+                    "links": links})
+    return out
+
+
+def _state(links: list[dict]) -> tuple[str, str]:
+    return (NOT_CHECKED, "declared") if links else (NOT_LINKED, "unspecified")
+
+
+def write_paper_site(site: Path, paper: dict, claims: list[dict],
+                     page_size: list | tuple | None, pdf_url: str | None,
+                     web_dist: Path | None = None) -> dict:
+    """Write the paper page's data under `site/data/` (and the reader, given its built
+    directory). `paper` is the paper's registry entry (its identifier, title, authors, links and
+    formalisations), which the reader's panel beside the paper shows; `claims` are from
+    `paper_claims`. Replaces whatever is at `site`."""
+    from .pipeline import write_preview
+    projects = [f["repository"] for f in paper.get("formalisations", [])]
+    index_claims, chunks = [], {}
+    for c in claims:
+        cid = f"P:{c['key']}"
+        fname = mf.claim_file(cid)
+        anchor = {**c["anchor"], "kind": "approximate"} if c.get("anchor") else None
+        by_project = {}
+        for r in projects:
+            status, cls = _state([l for l in c["links"] if l["repository"] == r])
+            by_project[r] = {"status": status, "class": cls}
+        status, cls = _state([l for l in c["links"]
+                              if not projects or l["repository"] == projects[0]])
+        common = {"id": cid, "paper": "P", "number": c["number"], "title": c.get("title"),
+                  "heading": c.get("title") or None, "kind": c["kind"], "scope": "formal",
+                  "status": status, "class": cls, "by_project": by_project}
+        index_claims.append({**common, "anchor": mf.index_anchor(anchor, "approximate"),
+                             "counts": {}, "file": fname, "linked": len(c["links"]),
+                             "suggested": 0})
+        chunks[fname] = {
+            **common, "schema": SCHEMA, "mode": "simple", "release": "paper", "key": c["key"],
+            "scope_note": "", "scope_reviewed": True, "anchor": anchor,
+            "source_matches": None, "statement_tex": None, "blockers": [],
+            "links": [{**l, "status": NOT_CHECKED, "class": "declared", "why": None,
+                       "statement": None, "path": None} for l in c["links"]],
+            "suggestions": [], "clauses": [], "decls": {}, "findings": []}
+    summary = {}
+    for c in index_claims:
+        summary[c["status"]] = summary.get(c["status"], 0) + 1
+    # The page has no Lean evidence; its identifier is that of what it shows.
+    eid = content_id({"paper": paper, "claims": claims, "pdf": pdf_url})
+    index = {"schema": SCHEMA, "mode": "paper", "release": "paper", "evidence": eid,
+             "tool": {"name": "machinechecked"}, "project": {}, "projects": projects,
+             "paper": paper, "notes": [],
+             "papers": [{"id": "P", "title": paper.get("title", ""), "mode": "pdf", "pdf": None,
+                         "pdf_url": pdf_url, "page_size": list(page_size) if page_size else None,
+                         "note": ""}],
+             "summary": {"P": summary},
+             "classes": {c["status"]: c["class"] for c in index_claims},
+             "findings": {"total": {}, "general": []}, "gates": {}, "claims": index_claims,
+             "search": "data/search.json"}
+    write_preview(site, index, chunks, {}, {}, web_dist)
+    return index

{machinechecked-0.1.3 → machinechecked-0.1.4}/engine/python/prooflens/simple.py

@@ -679,16 +679,11 @@ def write_view_site(view: dict, site: Path, web_dist: Path | None,
                        if (a.get("label") or f"{a['env']}-{a.get('number')}") == c["key"]), None)
         fname = mf.claim_file(c["id"])
         akind = "approximate" if anchor and anchor.get("kind") == "pdf" else "exact"
-        top = mf.top_of({**anchor, "kind": akind}) if anchor and anchor.get("begin") else None
         index_claims.append({
             "id": c["id"], "paper": c["paper"], "number": c["number"], "title": c["title"],
             "heading": c["title"] or None, "kind": c["kind"], "scope": c["scope"],
             "status": c["status"], "class": c["class"],
-            "anchor": {"kind": akind, "page": top[0], "y0": top[1],
-                       "y1": anchor["end"][2] if anchor.get("end") and
-                       anchor["end"][0] == anchor["begin"][0] else None,
-                       "page_end": anchor["end"][0] if anchor.get("end") else None}
-            if top else None,
+            "anchor": mf.index_anchor(anchor, akind),
             "counts": {}, "file": fname, "linked": len(c["links"]),
             "suggested": len(c.get("suggestions", []))})
         names = set()