Documenti d'Italia v1.0: 204 gold annotations, schema, scorer, inaugural leaderboard
6bdf6a4 verified | #!/usr/bin/env python3 | |
| """Fetch the non-rehosted documents of Documenti d'Italia from their original | |
| public sources, verifying sha256. Stdlib only. Re-run to resume; already-valid | |
| files are skipped. Files that disappear from the web are reported — their | |
| annotations remain usable and scoreable for the subset you hold.""" | |
| import hashlib, json, os, sys, time, urllib.request | |
| HERE = os.path.dirname(os.path.abspath(__file__)) | |
| UA = "documenti-italia-fetch/1.0 (dataset reconstruction)" | |
| ok = miss = skip = 0 | |
| for line in open(os.path.join(HERE, "manifest.jsonl")): | |
| e = json.loads(line) | |
| if e.get("file"): | |
| continue | |
| if not e.get("url") or e["url"].startswith("unknown"): | |
| print("UNAVAILABLE (no known source):", e["doc_id"][:16]); miss += 1; continue | |
| dst = os.path.join(HERE, "docs", "fetched", e["sha256"] + "." + e["format"]) | |
| if os.path.exists(dst) and hashlib.sha256(open(dst, "rb").read()).hexdigest() == e["sha256"]: | |
| skip += 1; continue | |
| os.makedirs(os.path.dirname(dst), exist_ok=True) | |
| try: | |
| req = urllib.request.Request(e["url"], headers={"User-Agent": UA}) | |
| data = urllib.request.urlopen(req, timeout=60).read() | |
| if hashlib.sha256(data).hexdigest() != e["sha256"]: | |
| print("CHANGED upstream:", e["url"][:80]); miss += 1; continue | |
| open(dst, "wb").write(data); ok += 1; print("ok", e["sha256"][:12]) | |
| time.sleep(1) | |
| except Exception as ex: | |
| print("FAILED:", e["url"][:80], "-", str(ex)[:60]); miss += 1 | |
| print(f"fetched {ok}, skipped {skip}, unavailable {miss}") | |