#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""作业 5 · 指标解析自测（不连 GPU，先红后绿）。"""
from __future__ import annotations

import sys

import _bootstrap  # noqa: F401
from 作业5_vllm_health import (
    parse_docker_status,
    parse_nvidia_smi,
    parse_vllm_queue,
    get_gpu_info,
    get_request_queue,
    get_vllm_status,
)


def test_parse_docker_status() -> None:
    raw = "/vllm-qwen3.6-27b|true|running|2026-08-18T01:00:00Z|vllm/vllm-openai:latest"
    got = parse_docker_status(raw)
    assert got["running"] is True
    assert got["status"] == "running"
    assert got["container"] == "vllm-qwen3.6-27b"


def test_parse_nvidia_smi() -> None:
    raw = "0, NVIDIA H20, 37, 0 %, 83868 MiB, 97871 MiB"
    got = parse_nvidia_smi(raw)
    assert got["gpus"][0]["name"] == "NVIDIA H20"
    assert got["gpus"][0]["temperature_c"] == 37
    assert got["gpus"][0]["utilization"] == "0 %"
    assert got["gpus"][0]["memory_used_mib"] == 83868
    assert got["gpus"][0]["memory_total_mib"] == 97871


def test_parse_vllm_queue() -> None:
    raw = """
# HELP vllm:num_requests_waiting ...
vllm:num_requests_running{engine="0",model_name="qwen3.6-27b"} 1.0
vllm:num_requests_waiting{engine="0",model_name="qwen3.6-27b"} 3.0
vllm:num_requests_waiting_by_reason{engine="0",model_name="qwen3.6-27b",reason="capacity"} 2.0
vllm:kv_cache_usage_perc{engine="0",model_name="qwen3.6-27b"} 0.42
"""
    got = parse_vllm_queue(raw)
    assert got["waiting"] == 3.0
    assert got["running"] == 1.0
    assert got["kv_cache_usage_perc"] == 0.42


def test_tools_use_injected_runner() -> None:
    def runner(cmd: str) -> str:
        if "inspect" in cmd:
            return "/vllm-qwen3.6-27b|true|running|t|img"
        if "nvidia-smi" in cmd:
            return "0, NVIDIA H20, 37, 0 %, 100 MiB, 200 MiB"
        if "metrics" in cmd:
            return 'vllm:num_requests_waiting{engine="0"} 5.0\nvllm:num_requests_running{engine="0"} 1.0\n'
        raise AssertionError(cmd)

    st = get_vllm_status(runner=runner)
    gpu = get_gpu_info(runner=runner)
    q = get_request_queue(runner=runner)
    assert st["running"] is True
    assert gpu["gpus"][0]["name"] == "NVIDIA H20"
    assert q["waiting"] == 5.0


def main() -> int:
    test_parse_docker_status()
    test_parse_nvidia_smi()
    test_parse_vllm_queue()
    test_tools_use_injected_runner()
    print("[PASS] 作业5 解析自测")
    return 0


if __name__ == "__main__":
    sys.exit(main())