Patchnoisseur / dataset /cve_diff /validate_diffs.py
michoo42's picture
dataset
c4c44be
#!/usr/bin/env python3
"""Validate that every file under data/ contains real diff content.
A file is valid when, for each `=== <url> ===` section (or for the whole
body if there is no separator), the section contains `diff --git ` at
column 0 OR starts with `From <sha40> Mon Sep 17 00:00:00 2001`.
Invalid files and the reason are written to .logs/invalid_diffs.tsv and
printed as a summary on stdout.
Usage:
python validate_diffs.py [DATA_DIR]
If DATA_DIR is omitted, defaults to ./data next to this script.
"""
from __future__ import annotations
import pathlib
import re
import sys
from collections import Counter
SCRIPT_DIR = pathlib.Path(__file__).resolve().parent
DEFAULT_DATA_DIR = SCRIPT_DIR / "data"
LOG_DIR = SCRIPT_DIR / ".logs"
_DIFF_GIT_RE = re.compile(rb"^diff --git ", re.M)
_GIT_FORMAT_PATCH_RE = re.compile(rb"^From [0-9a-f]{40} Mon Sep ")
_SECTION_SEP_RE = re.compile(rb"^=== (.+?) ===$", re.M)
_HTML_HINT_RE = re.compile(rb"<(?:html|head|body|!doctype)\b", re.I)
def looks_like_diff(body: bytes) -> bool:
if not body:
return False
if _DIFF_GIT_RE.search(body):
return True
return bool(_GIT_FORMAT_PATCH_RE.match(body.split(b"\n", 1)[0]))
def classify_bad(body: bytes) -> str:
"""Return a short reason describing why `body` is not a diff."""
if not body:
return "empty file"
stripped = body.strip()
if not stripped:
return "whitespace only"
head = body[:4096]
if _HTML_HINT_RE.search(head):
title = re.search(rb"<title[^>]*>(.*?)</title>", head, re.I | re.S)
if title:
t = title.group(1).strip().decode("utf-8", "replace")[:120]
return f"HTML page (title: {t!r})"
return "HTML page"
if head.lstrip().startswith((b"{", b"[")):
return "JSON payload (likely API error)"
if b"Not Found" in head or b"404" in head[:200]:
return "looks like 404 / not-found text"
if b"Rate limit" in head or b"rate limit" in head:
return "rate-limit message"
try:
text_preview = head.decode("utf-8")
except UnicodeDecodeError:
return f"binary / non-utf8 content ({len(body)} B)"
snippet = " ".join(text_preview.split())[:120]
return f"no `diff --git` and no git-format-patch header; starts with: {snippet!r}"
def validate_file(path: pathlib.Path) -> list[str]:
"""Return a list of reasons why `path` is invalid (empty list = OK)."""
try:
body = path.read_bytes()
except OSError as e:
return [f"unreadable: {e}"]
if not body:
return ["empty file"]
sep_matches = list(_SECTION_SEP_RE.finditer(body))
if not sep_matches:
if looks_like_diff(body):
return []
return [classify_bad(body)]
reasons: list[str] = []
for i, m in enumerate(sep_matches):
url = m.group(1).decode("utf-8", "replace")
start = m.end() + 1 # skip newline after separator
end = sep_matches[i + 1].start() if i + 1 < len(sep_matches) else len(body)
section = body[start:end]
if not looks_like_diff(section):
reasons.append(f"section {i + 1} ({url}): {classify_bad(section)}")
return reasons
def main() -> int:
data_dir = pathlib.Path(sys.argv[1]) if len(sys.argv) > 1 else DEFAULT_DATA_DIR
if not data_dir.is_dir():
print(f"error: {data_dir} is not a directory", file=sys.stderr)
return 2
LOG_DIR.mkdir(parents=True, exist_ok=True)
log_path = LOG_DIR / "invalid_diffs.tsv"
files = sorted(p for p in data_dir.rglob("*") if p.is_file())
print(f"[scan] {len(files)} file(s) under {data_dir}", file=sys.stderr)
invalid = 0
reason_counts: Counter[str] = Counter()
with open(log_path, "w", encoding="utf-8") as log:
log.write("path\treason\n")
for path in files:
reasons = validate_file(path)
if not reasons:
continue
invalid += 1
rel = path.relative_to(data_dir)
for r in reasons:
# bucket the reason for the summary (keep the prefix)
head = r.split(":", 1)[0].split(";", 1)[0][:80]
reason_counts[head] += 1
log.write(f"{rel}\t{r}\n")
ok = len(files) - invalid
print(f"[done] ok={ok} invalid={invalid} log={log_path}", file=sys.stderr)
if reason_counts:
print("[summary] top reasons:", file=sys.stderr)
for reason, n in reason_counts.most_common(10):
print(f" {n:>6} {reason}", file=sys.stderr)
return 1 if invalid else 0
if __name__ == "__main__":
sys.exit(main())