metis/scripts/publish_test_metrics.py

#!/usr/bin/env python3
"""Publish Metis quality-gate test metrics to Pushgateway (Prometheus ingest)."""

from __future__ import annotations

import json
import os
import sys
import urllib.request
import xml.etree.ElementTree as ET


def _escape_label(value: str) -> str:
    return value.replace("\\", "\\\\").replace("\n", "\\n").replace('"', '\\"')


def _label_str(labels: dict[str, str]) -> str:
    parts = [f'{key}="{_escape_label(val)}"' for key, val in labels.items() if val]
    return "{" + ",".join(parts) + "}" if parts else ""


def _load_coverage(path: str) -> float:
    with open(path, "r", encoding="utf-8") as handle:
        payload = json.load(handle)
    summary = payload.get("summary") or {}
    percent = summary.get("percent_covered")
    if isinstance(percent, (int, float)):
        return float(percent)
    raise RuntimeError("coverage summary missing percent_covered")


def _load_junit(path: str) -> dict[str, int]:
    tree = ET.parse(path)
    root = tree.getroot()

    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

    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 _read_http(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 _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"metrics push failed status={resp.status}")


def _fetch_existing_counter(pushgateway_url: str, metric: str, labels: dict[str, str]) -> float:
    text = _read_http(f"{pushgateway_url.rstrip('/')}/metrics")
    if not text:
        return 0.0

    for line in text.splitlines():
        if not line.startswith(metric + "{"):
            continue
        if any(f'{k}="{v}"' not in line for k, v in labels.items()):
            continue
        parts = line.split()
        if len(parts) < 2:
            continue
        try:
            return float(parts[1])
        except ValueError:
            return 0.0
    return 0.0


def main() -> int:
    coverage_path = os.getenv("COVERAGE_JSON", "build/coverage.json")
    junit_path = os.getenv("JUNIT_XML", "build/junit.xml")
    pushgateway_url = os.getenv(
        "PUSHGATEWAY_URL", "http://platform-quality-gateway.monitoring.svc.cluster.local:9091"
    ).strip()
    suite = os.getenv("SUITE_NAME", "metis")
    branch = os.getenv("BRANCH_NAME", "")
    build_number = os.getenv("BUILD_NUMBER", "")
    commit = os.getenv("GIT_COMMIT", "")

    if not os.path.exists(coverage_path):
        raise RuntimeError(f"missing coverage file {coverage_path}")
    if not os.path.exists(junit_path):
        raise RuntimeError(f"missing junit file {junit_path}")

    coverage = _load_coverage(coverage_path)
    totals = _load_junit(junit_path)
    passed = max(totals["tests"] - totals["failures"] - totals["errors"] - totals["skipped"], 0)

    outcome = "ok"
    if totals["tests"] <= 0 or totals["failures"] > 0 or totals["errors"] > 0:
        outcome = "failed"

    job_name = "platform-quality-ci"
    ok_count = _fetch_existing_counter(
        pushgateway_url,
        "platform_quality_gate_runs_total",
        {"job": job_name, "suite": suite, "status": "ok"},
    )
    failed_count = _fetch_existing_counter(
        pushgateway_url,
        "platform_quality_gate_runs_total",
        {"job": job_name, "suite": suite, "status": "failed"},
    )
    if outcome == "ok":
        ok_count += 1
    else:
        failed_count += 1

    labels = {
        "suite": suite,
        "branch": branch,
        "build_number": build_number,
        "commit": commit,
    }
    payload_lines = [
        "# 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 metis_quality_gate_tests_total gauge",
        f'metis_quality_gate_tests_total{{suite="{suite}",result="passed"}} {passed}',
        f'metis_quality_gate_tests_total{{suite="{suite}",result="failed"}} {totals["failures"]}',
        f'metis_quality_gate_tests_total{{suite="{suite}",result="error"}} {totals["errors"]}',
        f'metis_quality_gate_tests_total{{suite="{suite}",result="skipped"}} {totals["skipped"]}',
        "# TYPE metis_quality_gate_coverage_percent gauge",
        f'metis_quality_gate_coverage_percent{{suite="{suite}"}} {coverage:.3f}',
        "# TYPE metis_quality_gate_build_info gauge",
        f"metis_quality_gate_build_info{_label_str(labels)} 1",
    ]
    payload = "\n".join(payload_lines) + "\n"
    _post_text(f"{pushgateway_url.rstrip('/')}/metrics/job/{job_name}/suite/{suite}", payload)

    print(
        json.dumps(
            {
                "suite": suite,
                "outcome": outcome,
                "tests_total": totals["tests"],
                "tests_passed": passed,
                "tests_failed": totals["failures"],
                "tests_errors": totals["errors"],
                "tests_skipped": totals["skipped"],
                "coverage_percent": round(coverage, 3),
                "ok_counter": ok_count,
                "failed_counter": failed_count,
            },
            indent=2,
        )
    )
    return 0


if __name__ == "__main__":
    try:
        raise SystemExit(main())
    except Exception as exc:
        print(f"metrics push failed: {exc}")
        raise