Download docs/results/check_explained.py from NicolasT08/formal-mas: direct link, hf CLI and curl.
- Browser
- Download file 4.8 kB
-
https://huggingface.co/NicolasT08/formal-mas/resolve/main/docs/results/check_explained.py
- Command line
-
hf download hf://NicolasT08/formal-mas/docs/results/check_explained.py
-
curl -L -o check_explained.py https://huggingface.co/NicolasT08/formal-mas/resolve/main/docs/results/check_explained.py
4.8 kB
| """Check that every number in docs/results_explained.md matches docs/results/results.md. | |
| Usage (from the repository root, after `make_tables.py`): | |
| .venv/Scripts/python.exe docs/results/check_explained.py | |
| results.md is generated by make_tables.py from runs/; results_explained.md is hand-written around | |
| copies of its tables. The checks: | |
| 1. Generated tables. A Markdown table in results_explained.md whose header row appears in | |
| results.md must appear there verbatim, every row in the same order (a single block). | |
| 2. Notes. Every `*Note.* ...` paragraph must appear verbatim in results.md. | |
| 3. Every other number (prose and hand-written tables) must equal a number in results.md | |
| rounded to the same decimals (so "1.47" matches 1.4744, "174" matches 173.9). Skipped: | |
| headings, inline code, section references (§5.3, §7.1–7.3), dates (2026-11-08), link targets | |
| and digits that are part of a name (C5, H1, p6). | |
| Limitation: rule 3 checks that a number exists in results.md, not that it is the right one in | |
| context. Decimals rarely match by chance; small integers (counts) almost always do, so for them | |
| the check only catches numbers that appear nowhere in results.md. | |
| Exit code 0 if everything matches, 1 otherwise (each mismatch is printed with its line number). | |
| Reads only; writes nothing. | |
| """ | |
| from __future__ import annotations | |
| import re | |
| import sys | |
| from pathlib import Path | |
| HERE = Path(__file__).resolve().parent | |
| RESULTS = HERE / "results.md" | |
| EXPLAINED = HERE.parent / "results_explained.md" | |
| NUMBER = re.compile(r"(?<![\w.\-−])\d+(?:\.\d+)?(?![\w])") | |
| SKIP = [ | |
| re.compile(r"`[^`]*`"), # inline code (ids, paths, labels) | |
| re.compile(r"\]\([^)]*\)"), # link targets | |
| re.compile(r"§\s?\d+(?:\.\d+)?(?:\s?[–-]\s?\d+(?:\.\d+)?)?"), # section references and ranges | |
| re.compile(r"\d{4}-\d{2}-\d{2}(?:T[\d:.]+Z?)?"), # dates and timestamps | |
| re.compile(r"\d{2}:\d{2}(?::\d{2})?"), # clock times | |
| ] | |
| def table_blocks(lines: list[str]) -> list[tuple[int, list[str]]]: | |
| """(1-based start line, lines) of each Markdown table.""" | |
| blocks, cur, start = [], [], 0 | |
| for i, line in enumerate(lines, 1): | |
| if line.startswith("|"): | |
| if not cur: | |
| start = i | |
| cur.append(line) | |
| elif cur: | |
| blocks.append((start, cur)) | |
| cur = [] | |
| if cur: | |
| blocks.append((start, cur)) | |
| return blocks | |
| def decimals(tok: str) -> int: | |
| return len(tok.split(".")[1]) if "." in tok else 0 | |
| def main() -> int: | |
| res = RESULTS.read_text(encoding="utf-8") | |
| exp_lines = EXPLAINED.read_text(encoding="utf-8").splitlines() | |
| res_lines = res.splitlines() | |
| res_values = sorted({float(t) for t in re.findall(r"\d+(?:\.\d+)?", res)}) | |
| res_headers = {line for line in res_lines if line.startswith("|")} | |
| errors: list[str] = [] | |
| counts = {"tables": 0, "notes": 0, "numbers": 0} | |
| generated_lines: set[int] = set() | |
| for start, block in table_blocks(exp_lines): | |
| if block[0] not in res_headers: | |
| continue # hand-written table: its numbers go through rule 3 | |
| counts["tables"] += 1 | |
| generated_lines.update(range(start, start + len(block))) | |
| if "\n".join(block) not in res: | |
| bad = [f"line {start + k}: {row[:100]}" for k, row in enumerate(block) if row not in res_headers] | |
| errors.append(f"table at line {start} is not a verbatim block of results.md" | |
| + ("".join(f"\n {b}" for b in bad) if bad else " (rows reordered or missing)")) | |
| for i, line in enumerate(exp_lines, 1): | |
| if i in generated_lines or line.startswith("#"): | |
| continue | |
| if line.startswith("*Note.*"): | |
| counts["notes"] += 1 | |
| if line not in res_lines: | |
| errors.append(f"line {i}: note not found verbatim in results.md: {line[:100]}") | |
| continue | |
| text = line | |
| for pat in SKIP: | |
| text = pat.sub(" ", text) | |
| for m in NUMBER.finditer(text): | |
| tok = m.group() | |
| counts["numbers"] += 1 | |
| d, v = decimals(tok), float(tok) | |
| if not any(abs(round(x, d) - v) < 1e-9 for x in res_values): | |
| errors.append(f"line {i}: {tok} not in results.md (at {d} decimals): {line.strip()[:100]}") | |
| print(f"checked {counts['tables']} generated tables, {counts['notes']} notes and {counts['numbers']} other numbers " | |
| f"in {EXPLAINED.name} against {RESULTS.name}") | |
| if errors: | |
| print(f"FAIL: {len(errors)} mismatch(es)") | |
| for e in errors: | |
| print(f" - {e}") | |
| return 1 | |
| print("OK: every number matches") | |
| return 0 | |
| if __name__ == "__main__": | |
| sys.exit(main()) | |