atlasbot/scripts/publish_test_metrics.py

#!/usr/bin/env python3
"""Publish Atlasbot CI test metrics to Pushgateway.

Inputs:
- JUnit XML file and coverage JSON file.

Outputs:
- platform_quality_gate_runs_total{suite="atlasbot",status="ok|failed"}
- atlasbot_quality_gate_tests_total{suite="atlasbot",result=*}
- atlasbot_quality_gate_coverage_percent{suite="atlasbot"}
- platform_quality_gate_workspace_line_coverage_percent{suite="atlasbot"}
- platform_quality_gate_source_lines_over_500_total{suite="atlasbot"}
"""

from __future__ import annotations

import json
import os
import urllib.request
import xml.etree.ElementTree as ET
from pathlib import Path


def _as_int(node: ET.Element, name: str) -> int:
    raw = node.attrib.get(name) or "0"
    try:
        return int(float(raw))
    except ValueError:
        return 0


def _load_junit(path: Path) -> dict[str, int]:
    if not path.exists():
        return {"tests": 0, "failures": 0, "errors": 0, "skipped": 0}

    tree = ET.parse(path)
    root = tree.getroot()
    suites: list[ET.Element]
    if root.tag == "testsuite":
        suites = [root]
    elif root.tag == "testsuites":
        suites = list(root.findall("testsuite"))
    else:
        suites = []

    totals = {"tests": 0, "failures": 0, "errors": 0, "skipped": 0}
    for suite in suites:
        totals["tests"] += _as_int(suite, "tests")
        totals["failures"] += _as_int(suite, "failures")
        totals["errors"] += _as_int(suite, "errors")
        totals["skipped"] += _as_int(suite, "skipped")
    return totals


def _load_coverage_percent(path: Path) -> float:
    if not path.exists():
        return 0.0
    payload = json.loads(path.read_text(encoding="utf-8"))
    summary = payload.get("summary") or {}
    percent = summary.get("percent_covered")
    if isinstance(percent, (int, float)):
        return float(percent)
    return 0.0


def _count_source_lines_over_500(root: Path) -> int:
    if not root.exists():
        return 0
    over = 0
    for path in root.rglob("*.py"):
        if not path.is_file():
            continue
        line_count = sum(1 for _ in path.open("r", encoding="utf-8"))
        if line_count > 500:
            over += 1
    return over


def _read_text(url: str) -> str:
    try:
        with urllib.request.urlopen(url, timeout=10) as resp:
            return resp.read().decode("utf-8", errors="replace")
    except Exception:
        return ""


def _counter(metrics: str, suite: str, status: str) -> float:
    for line in metrics.splitlines():
        if not line.startswith("platform_quality_gate_runs_total{"):
            continue
        if f'job="platform-quality-ci"' not in line:
            continue
        if f'suite="{suite}"' not in line:
            continue
        if f'status="{status}"' not in line:
            continue
        parts = line.split()
        if len(parts) < 2:
            continue
        try:
            return float(parts[1])
        except ValueError:
            return 0.0
    return 0.0


def _post_text(url: str, payload: str) -> None:
    req = urllib.request.Request(
        url,
        data=payload.encode("utf-8"),
        method="POST",
        headers={"Content-Type": "text/plain"},
    )
    with urllib.request.urlopen(req, timeout=10) as resp:
        if resp.status >= 400:
            raise RuntimeError(f"push failed status={resp.status}")


def main() -> int:
    suite = os.getenv("SUITE_NAME", "atlasbot")
    pushgateway_url = os.getenv(
        "PUSHGATEWAY_URL", "http://platform-quality-gateway.monitoring.svc.cluster.local:9091"
    ).rstrip("/")

    junit_path = Path(os.getenv("JUNIT_PATH", "build/junit.xml"))
    coverage_path = Path(os.getenv("COVERAGE_PATH", "build/coverage.json"))
    source_root = Path(os.getenv("SOURCE_ROOT", "atlasbot"))

    totals = _load_junit(junit_path)
    coverage_pct = _load_coverage_percent(coverage_path)
    source_lines_over_500 = _count_source_lines_over_500(source_root)
    passed = max(totals["tests"] - totals["failures"] - totals["errors"] - totals["skipped"], 0)
    outcome = "ok" if totals["tests"] > 0 and totals["failures"] == 0 and totals["errors"] == 0 else "failed"

    metrics = _read_text(f"{pushgateway_url}/metrics")
    ok_count = _counter(metrics, suite, "ok")
    failed_count = _counter(metrics, suite, "failed")
    if outcome == "ok":
        ok_count += 1
    else:
        failed_count += 1

    payload = "\n".join(
        [
            "# TYPE platform_quality_gate_runs_total counter",
            f'platform_quality_gate_runs_total{{suite="{suite}",status="ok"}} {ok_count:.0f}',
            f'platform_quality_gate_runs_total{{suite="{suite}",status="failed"}} {failed_count:.0f}',
            "# TYPE atlasbot_quality_gate_tests_total gauge",
            f'atlasbot_quality_gate_tests_total{{suite="{suite}",result="passed"}} {passed}',
            f'atlasbot_quality_gate_tests_total{{suite="{suite}",result="failed"}} {totals["failures"]}',
            f'atlasbot_quality_gate_tests_total{{suite="{suite}",result="error"}} {totals["errors"]}',
            f'atlasbot_quality_gate_tests_total{{suite="{suite}",result="skipped"}} {totals["skipped"]}',
            "# TYPE atlasbot_quality_gate_coverage_percent gauge",
            f'atlasbot_quality_gate_coverage_percent{{suite="{suite}"}} {coverage_pct:.3f}',
            "# TYPE platform_quality_gate_workspace_line_coverage_percent gauge",
            f'platform_quality_gate_workspace_line_coverage_percent{{suite="{suite}"}} {coverage_pct:.3f}',
            "# TYPE platform_quality_gate_source_lines_over_500_total gauge",
            f'platform_quality_gate_source_lines_over_500_total{{suite="{suite}"}} {source_lines_over_500}',
        ]
    ) + "\n"

    _post_text(f"{pushgateway_url}/metrics/job/platform-quality-ci/suite/{suite}", payload)
    return 0


if __name__ == "__main__":
    raise SystemExit(main())