acra-backend / code /metrics_export.py
Zeref8425's picture
feat(metrics): export stored results with ownership and expiry checks
f52e861
Raw History Blame Contribute Delete
5.05 kB
"""Flatten stored job and test-run metrics into one export row.
Every image is kept. A failed pass flag or a no-conflict image is a row,
not a reason to drop it.
"""
from __future__ import annotations
import csv
import io
import json
from typing import Any, Dict, Iterable, List, Optional
COLUMNS = [
"source",
"id",
"filename",
"created_at",
"cvd_type",
"severity",
"de_before",
"de_after",
"de_improvement",
"conflicts_found",
"conflicts_resolved",
"conflict_resolution_rate",
"new_conflicts",
"unresolved_conflicts",
"new_conflict_pairs",
"unresolved_pairs",
"naturalness",
"no_conflict",
"pass_de",
"pass_resolution",
"pass_naturalness",
]
def _metrics(value: Any) -> dict:
if isinstance(value, str):
try:
parsed = json.loads(value)
except (TypeError, ValueError):
return {}
return parsed if isinstance(parsed, dict) else {}
return dict(value) if isinstance(value, dict) else {}
def _num(value: Any, digits: int = 6) -> Optional[float]:
if value is None or value == "":
return None
try:
return round(float(value), digits)
except (TypeError, ValueError):
return None
def _int(value: Any, default: int = 0) -> int:
if value is None or value == "":
return default
try:
return int(value)
except (TypeError, ValueError):
return default
def _pairs(raw: Any) -> List[List[int]]:
out = []
for pair in raw or []:
if isinstance(pair, (list, tuple)) and len(pair) == 2:
out.append([_int(pair[0]), _int(pair[1])])
return out
def _pair_text(pairs: List[List[int]]) -> str:
return ";".join(f"{a}-{b}" for a, b in pairs)
def _flag(metrics: dict, metrics_key: str, row: dict, row_key: str) -> bool:
if metrics_key in metrics:
return bool(metrics.get(metrics_key))
if row_key in row and row.get(row_key) is not None:
return bool(row.get(row_key))
return False
def metric_row(source: str, row: dict) -> dict:
"""One spreadsheet row. Caller decides which records to pass; this never filters."""
metrics = _metrics(row.get("metrics"))
diag = metrics.get("conflict_diagnostics") or {}
if not isinstance(diag, dict):
diag = {}
found = metrics.get("conflicts_found")
if found is None:
found = metrics.get("n_conflicts_total")
found_n = _int(found, 0)
resolved = metrics.get("n_conflicts_resolved")
resolved_pairs = _pairs(diag.get("resolved_pairs"))
if resolved is None:
resolved_n = len(resolved_pairs)
else:
resolved_n = _int(resolved, 0)
new_pairs = _pairs(diag.get("new_pairs"))
unresolved_pairs = _pairs(diag.get("unresolved_pairs"))
no_conflict = bool(metrics.get("already_accessible")) or found_n == 0
record_id = row.get("job_id") or row.get("run_id") or row.get("id") or ""
return {
"source": source,
"id": str(record_id),
"filename": row.get("filename") or "",
"created_at": row.get("created_at") or "",
"cvd_type": row.get("cvd_type") or "",
"severity": _num(row.get("severity")),
"de_before": _num(metrics.get("de_before_mean")),
"de_after": _num(metrics.get("de_after_mean")),
"de_improvement": _num(
metrics.get("delta_e_improvement", metrics.get("de_improvement"))
),
"conflicts_found": found_n,
"conflicts_resolved": resolved_n,
"conflict_resolution_rate": _num(metrics.get("conflict_resolution_rate")),
"new_conflicts": len(new_pairs),
"unresolved_conflicts": len(unresolved_pairs),
"new_conflict_pairs": _pair_text(new_pairs),
"unresolved_pairs": _pair_text(unresolved_pairs),
"naturalness": _num(metrics.get("naturalness_preservation")),
"no_conflict": no_conflict,
"pass_de": _flag(metrics, "pass_de_improvement", row, "pass_de"),
"pass_resolution": _flag(metrics, "pass_resolution_rate", row, "pass_resolution"),
"pass_naturalness": _flag(metrics, "pass_naturalness", row, "pass_naturalness"),
}
def to_csv(rows: Iterable[dict]) -> str:
buf = io.StringIO()
writer = csv.DictWriter(buf, fieldnames=COLUMNS, lineterminator="\n", extrasaction="ignore")
writer.writeheader()
for row in rows:
flat = {}
for key in COLUMNS:
value = row.get(key, "")
if isinstance(value, bool):
flat[key] = "true" if value else "false"
elif value is None:
flat[key] = ""
elif isinstance(value, str) and value.lstrip().startswith(("=", "+", "-", "@")):
# Uploaded filenames must remain text when opened in a spreadsheet.
flat[key] = "'" + value
else:
flat[key] = value
writer.writerow(flat)
return buf.getvalue()
def to_json(rows: Iterable[dict]) -> str:
return json.dumps(list(rows), indent=2)