Return (summaries, site_series) for the process with the most dumps. Every dump lists every site, so the Nth occurrence of a site is its Nth sample; a site first seen late is padded with zeros so series align.
(path)
| 24 | |
| 25 | |
| 26 | def load_profile(path): |
| 27 | """Return (summaries, site_series) for the process with the most dumps. |
| 28 | Every dump lists every site, so the Nth occurrence of a site is its Nth |
| 29 | sample; a site first seen late is padded with zeros so series align.""" |
| 30 | procs = defaultdict(lambda: {"summaries": [], "series": defaultdict(list)}) |
| 31 | cur = None |
| 32 | with open(path, "r", errors="replace") as handle: |
| 33 | for line in handle: |
| 34 | line = line.strip() |
| 35 | if not line.startswith("{"): |
| 36 | continue |
| 37 | try: |
| 38 | record = json.loads(line) |
| 39 | except json.JSONDecodeError: |
| 40 | continue |
| 41 | if "memwaste" in record: |
| 42 | cur = procs[record.get("pid", 0)] |
| 43 | cur["summaries"].append(record) |
| 44 | elif cur is not None and "site" in record and "work" not in record: |
| 45 | samples = cur["series"][record["site"]] |
| 46 | while len(samples) < len(cur["summaries"]) - 1: |
| 47 | samples.append({"live_bytes": 0}) |
| 48 | samples.append(record) |
| 49 | if not procs: |
| 50 | return [], {} |
| 51 | best = max(procs.values(), key=lambda p: len(p["summaries"])) |
| 52 | return best["summaries"], best["series"] |
| 53 | |
| 54 | |
| 55 | def load_symbols(binary, summaries, sites): |