"""Web 层测试：API 契约、软删除语义、站点归一化、路径重写。

**为什么要有契约测试**：前端（``web/app.js``）与后端（``web/server.py``）是
两个独立文件，改动其中一个很容易忘记另一个。这里把「前端会调用的接口」
写成断言 —— 只要后端删掉某个路由，测试立刻失败，而不是等到页面上报错。
"""

from __future__ import annotations

import json
import re
from pathlib import Path

import pytest

from hdiv.core.paths import project_root
from hdiv.web import site
from hdiv.web.server import ROUTES


# ---------------------------------------------------------------------------
# 前端 ↔ 后端 接口契约
# ---------------------------------------------------------------------------

#: 前端实际会调用的接口（method, 路径样例）
#: 路径样例用于匹配路由正则；改前端时需同步此表。
FRONTEND_CALLS: list[tuple[str, str]] = [
    ("GET", "/api/health"),
    ("GET", "/api/config/display"),
    ("GET", "/api/summary"),
    ("GET", "/api/universes"),
    ("GET", "/api/universes/abc123"),
    ("GET", "/api/universes/abc123/members"),
    ("GET", "/api/universes/abc123/members/600519.SH"),
    ("GET", "/api/universes/abc123/backtests"),
    ("PATCH", "/api/universes/abc123"),
    ("GET", "/api/stocks/600519.SH"),
    ("GET", "/api/backtests"),
    ("GET", "/api/backtests/abc123"),
    ("GET", "/api/backtests/abc123/metrics"),
    ("GET", "/api/backtests/abc123/equity"),
    ("GET", "/api/backtests/abc123/trades"),
    # 净值曲线右轴可叠加的基准指数
    ("GET", "/api/indices"),
    ("GET", "/api/backtests/abc123/signals"),
    ("PATCH", "/api/backtests/abc123"),
    # 回测内分析：任意日持仓 + 个股买卖点
    ("GET", "/api/backtests/abc123/portfolio"),
    ("GET", "/api/backtests/abc123/position-dates"),
    ("GET", "/api/backtests/abc123/stocks"),
    ("GET", "/api/backtests/abc123/stocks/600519.SH"),
    # Walk-forward 样本外
    ("GET", "/api/walkforwards"),
    ("GET", "/api/walkforwards/abc123"),
]


def _match(method: str, path: str) -> bool:
    return any(m == method and pat.match(path) for m, pat, _ in ROUTES)


@pytest.mark.parametrize("method,path", FRONTEND_CALLS)
def test_frontend_api_calls_have_routes(method: str, path: str) -> None:
    assert _match(method, path), f"前端会调用 {method} {path}，但后端没有对应路由"


def test_app_js_only_calls_known_api_prefixes() -> None:
    """app.js 里出现的接口路径必须在契约表中有覆盖。"""
    js = (project_root() / "web" / "app.js").read_text(encoding="utf-8")
    # 抓取 api(`xxx`) / api('xxx') / patch(`xxx`, ...) 的首段
    calls = re.findall(r"\b(?:api|patch)\(\s*[`'\"]([a-z][\w/\-${}.\[\]]*)", js)
    assert calls, "未从 app.js 中解析出任何接口调用（解析逻辑需更新）"
    # FRONTEND_CALLS 里是绝对路径 /api/xxx；app.js 里的参数是相对 api 基址的
    # （api('summary') 实际请求 /api/summary），因此要去掉 api/ 前缀再比较。
    known = {
        p.removeprefix("/api/").strip("/").split("/")[0]
        for _m, p in FRONTEND_CALLS
    }
    unknown = {
        c.split("/")[0].split("?")[0]
        for c in calls
        if c.split("/")[0].split("?")[0] not in known
    }
    # 允许 ${...} 模板片段（这些是动态拼出的 run_id / symbol，前缀已在表中）
    unknown = {u for u in unknown if not u.startswith("$")}
    assert not unknown, f"app.js 调用了契约表未覆盖的接口前缀：{sorted(unknown)}"


def test_members_endpoint_defaults_to_selected() -> None:
    """接口层默认值：不传 passed 时应等价于 passed=1，而非全部候选。"""
    import inspect

    from hdiv.web import server

    src = inspect.getsource(server._members)
    assert 'passed_raw == "all"' in src or "passed_raw == 'all'" in src, \
        "默认值应只把显式的 all 当作「全部候选」"
    assert 'passed_raw != "0"' in src or "passed_raw != '0'" in src, \
        "未指定 passed 时应视为 1（仅入选）"


def test_every_route_is_covered_by_the_contract() -> None:
    """反向检查：每条后端路由都必须出现在契约表里。

    这样契约表就是「前后端接口清单」的唯一事实来源：
    新增接口忘了登记会被发现，删接口忘了清契约也会被发现。

    早期版本给两个「可选下钻」接口开了后门（阈值 <= 2），
    结果新增的三个接口漏登记却被放行 —— 所以现在零容忍。
    """
    known = {p for _m, p in FRONTEND_CALLS}
    uncovered = []
    for method, pat, _fn in ROUTES:
        if not any(m == method and pat.match(p) for m, p in FRONTEND_CALLS) and \
           not any(pat.match(p) for p in known):
            uncovered.append((method, pat.pattern))
    assert not uncovered, (
        f"以下路由未登记在 FRONTEND_CALLS 中：{uncovered}\n"
        "新增接口时请同步更新契约表，否则前端改动无法被发现。"
    )


# ---------------------------------------------------------------------------
# 资源路径重写（归档后图表不能失效）
# ---------------------------------------------------------------------------


def test_rewrite_asset_paths_depth0() -> None:
    html = '<script src="assets/echarts.min.js"></script>'
    assert site.rewrite_asset_paths(html, 0) == html


def test_rewrite_asset_paths_depth2() -> None:
    html = '<script src="assets/echarts.min.js"></script>'
    out = site.rewrite_asset_paths(html, 2)
    assert 'src="../../assets/echarts.min.js"' in out


def test_rewrite_does_not_touch_absolute_or_external() -> None:
    for html in (
        '<script src="/assets/echarts.min.js"></script>',
        '<script src="https://cdn.example.com/echarts.min.js"></script>',
        '<script src="//cdn.example.com/echarts.min.js"></script>',
        '<script src="../assets/echarts.min.js"></script>',
    ):
        assert site.rewrite_asset_paths(html, 2) == html, f"不该改写：{html}"


def test_rewrite_handles_href_too() -> None:
    html = '<link href="assets/app.css" rel="stylesheet">'
    assert '../../assets/app.css' in site.rewrite_asset_paths(html, 2)


# ---------------------------------------------------------------------------
# 前端文件
# ---------------------------------------------------------------------------


def test_frontend_source_files_exist() -> None:
    src = project_root() / "web"
    for name in ("index.html", "app.css", "app.js"):
        assert (src / name).is_file(), f"缺少前端文件 web/{name}"


def test_frontend_index_references_local_assets_only() -> None:
    html = (project_root() / "web" / "index.html").read_text(encoding="utf-8")
    refs = re.findall(r'(?:src|href)="([^"]+)"', html)
    external = [r for r in refs if r.startswith(("http://", "https://", "//"))]
    assert not external, f"前端引用了外部资源（违反离线约束）：{external}"
    assert "app/app.js" in refs and "app/app.css" in refs
    assert "assets/echarts.min.js" in refs


def test_frontend_uses_hash_routing_only() -> None:
    """hash 路由是刻意的：nginx 只需托管静态文件，不必配置 rewrite。"""
    js = (project_root() / "web" / "app.js").read_text(encoding="utf-8")
    assert "location.hash" in js
    # 不应出现 history.pushState 这类需要服务端配合的写法
    assert "pushState" not in js


def test_frontend_escapes_html() -> None:
    """用户可输入记录名称/备注，必须转义以避免 XSS。"""
    js = (project_root() / "web" / "app.js").read_text(encoding="utf-8")
    assert "function esc(" in js or "const esc =" in js
    assert "&quot;" in js and "&lt;" in js


# ---------------------------------------------------------------------------
# API 行为（需要数据库）
# ---------------------------------------------------------------------------


def _db_ready() -> bool:
    try:
        from hdiv.core.config import load_config
        from hdiv.data import db

        db.load_dotenv_once()
        db.list_tables(load_config("datasource"))
        return True
    except Exception:
        return False


requires_db = pytest.mark.skipif(not _db_ready(), reason="数据库不可用")


@requires_db
def test_service_summary_shape() -> None:
    from hdiv.web import service

    s = service.summary()
    for k in ("universes", "backtests", "profiles", "db"):
        assert k in s, f"summary 缺少字段 {k}"


@requires_db
def test_soft_delete_is_reversible_and_never_physical() -> None:
    """软删除语义：记录仍在库中，可通过 include_deleted 找回。"""
    from hdiv.web import service

    runs = service.list_universes()
    if not runs:
        pytest.skip("没有筛选记录可供测试")
    rid = runs[0]["run_id"]
    try:
        service.update_universe(rid, {"deleted": True})
        visible = {r["run_id"] for r in service.list_universes()}
        assert rid not in visible, "软删除后不应出现在默认列表"
        allruns = {r["run_id"] for r in service.list_universes(include_deleted=True)}
        assert rid in allruns, "软删除的记录必须仍可通过 include_deleted 找回（不做物理删除）"
    finally:
        service.update_universe(rid, {"deleted": False})
    assert rid in {r["run_id"] for r in service.list_universes()}


@requires_db
def test_archive_toggles_visibility() -> None:
    from hdiv.web import service

    runs = service.list_universes()
    if not runs:
        pytest.skip("没有筛选记录可供测试")
    rid = runs[0]["run_id"]
    try:
        service.update_universe(rid, {"archived": True})
        assert rid not in {r["run_id"] for r in service.list_universes()}
        assert rid in {r["run_id"] for r in service.list_universes(include_archived=True)}
    finally:
        service.update_universe(rid, {"archived": False})


@requires_db
def test_rename_persists_and_nullable() -> None:
    from hdiv.web import service

    runs = service.list_universes()
    if not runs:
        pytest.skip("没有筛选记录可供测试")
    rid = runs[0]["run_id"]
    original = runs[0]["display_name"]
    try:
        got = service.update_universe(rid, {"display_name": "单元测试名称"})
        assert got["display_name"] == "单元测试名称"
        assert got["title"] == "单元测试名称"
        # 清空后应回落到自动标题，而不是留空
        got = service.update_universe(rid, {"display_name": ""})
        assert got["display_name"] is None
        assert got["title"], "清空命名后应回落到自动标题"
    finally:
        service.update_universe(rid, {"display_name": original or ""})


@requires_db
def test_universe_backtest_link_is_bidirectional() -> None:
    from hdiv.web import service

    runs = service.list_universes()
    bts = service.list_backtests()
    if not runs or not bts:
        pytest.skip("缺少筛选记录或回测记录")
    rid, bid = runs[0]["run_id"], bts[0]["run_id"]
    original = bts[0]["universe_run_id"]
    try:
        service.update_backtest(bid, {"universe_run_id": rid})
        linked = service.list_backtests(universe_run_id=rid)
        assert bid in {x["run_id"] for x in linked}, "回测侧应能按股票池过滤到"
        u = service.get_universe(rid)
        assert u["linked_backtests"] >= 1, "股票池侧应统计到关联回测数"
    finally:
        service.update_backtest(bid, {"universe_run_id": original})


@requires_db
def test_link_rejects_nonexistent_universe() -> None:
    from hdiv.web import service

    bts = service.list_backtests()
    if not bts:
        pytest.skip("没有回测记录")
    with pytest.raises(ValueError):
        service.update_backtest(bts[0]["run_id"], {"universe_run_id": "不存在的runid"})


@requires_db
def test_update_rejects_empty_patch() -> None:
    from hdiv.web import service

    runs = service.list_universes()
    if not runs:
        pytest.skip("没有筛选记录")
    with pytest.raises(ValueError):
        service.update_universe(runs[0]["run_id"], {})


@requires_db
def test_all_api_payloads_are_json_serializable() -> None:
    """回归：pandas 的 numpy 标量曾导致接口 500。"""
    from hdiv.web import service
    from hdiv.web.server import _Encoder

    runs = service.list_universes()
    payloads: list[object] = [service.summary(), runs]
    if runs:
        rid = runs[0]["run_id"]
        payloads += [
            service.get_universe(rid),
            service.list_members(rid, size=2),
            service.list_backtests(universe_run_id=rid),
        ]
    bts = service.list_backtests()
    if bts:
        bid = bts[0]["run_id"]
        payloads += [
            service.get_backtest(bid),
            service.get_backtest_metrics(bid),
            service.get_backtest_equity(bid),
            service.list_backtest_trades(bid, size=2),
            service.list_backtest_signals(bid),
        ]
    for p in payloads:
        if p is None:
            continue
        json.dumps(p, ensure_ascii=False, cls=_Encoder)  # 不应抛异常


@requires_db
def test_equity_index_overlay_is_date_aligned() -> None:
    """净值曲线右轴叠加的指数必须与日期**逐点对齐**。

    类目轴上每个类目一个点：指数序列只要少一天，
    整条指数线就会相对净值曲线整体错位，画出错误的对比。
    """
    from hdiv.core.errors import HdivError
    from hdiv.web import service

    rid = _sample_backtest_run()
    if not rid:
        pytest.skip("没有可用的回测")
    items = service.list_indices()
    if not items:
        pytest.skip("hd_index_daily 没有指数行情")
    assert [x["code"] for x in items if x["is_default"]] == [service.DEFAULT_INDEX_CODE], \
        "应恰好把默认指数（沪深300）标成 default"

    base = service.get_backtest_equity(rid)
    if not base["dates"]:
        pytest.skip("该回测没有净值曲线")
    assert base["index"] is None, "不传 index 时不应凭空叠加指数"

    for it in items:
        ix = service.get_backtest_equity(rid, index_code=it["code"])["index"]
        assert ix["code"] == it["code"] and ix["name"]
        assert len(ix["close"]) == len(base["dates"]), f"{it['code']} 未与净值曲线对齐"
        vals = [v for v in ix["close"] if v is not None]
        # 叠加的是指数点位，不是净值；量级错了说明取错了列
        assert not vals or min(vals) > 10, f"{it['code']} 取值不像指数点位：{vals[:3]}"
        json.dumps(ix, allow_nan=False)

    # 库里没有的指数应当明确报错，而不是画一条空线
    with pytest.raises(HdivError):
        service.get_backtest_equity(rid, index_code="999999.XX")


@requires_db
def test_reason_text_is_human_readable() -> None:
    """成交理由必须渲染成人话，而不是丢一坨 JSON 给前端。"""
    from hdiv.web.service import _reason_text

    txt = _reason_text({
        "dividend_yield": 0.0575, "yield_percentile": 83.4,
        "rule": "股息率历史分位 83.4% >= P75", "observation_count": 1200,
    })
    assert "股息率 5.75%" in txt
    assert "历史分位 83.4%" in txt
    assert "{" not in txt and "}" not in txt


# ---------------------------------------------------------------------------
# 部署资产（缺了它们就会出现「API 不可用」）
# ---------------------------------------------------------------------------


def test_nginx_example_has_api_proxy_before_static() -> None:
    """回归：nginx 示例必须包含 /api 反代，且排在静态规则之前。

    漏掉反代块时，/ggx/api/health 会被当静态文件去 output/api/health 找 → 404，
    页面能打开但显示「API 不可用」。
    """
    cfg = (project_root() / "deploy" / "nginx.conf.example").read_text(encoding="utf-8")
    proxy = cfg.find("proxy_pass")
    assert proxy != -1, "nginx 示例缺少 proxy_pass（API 无法访问）"
    # 反代块必须出现在静态 location 之前（nginx 前缀匹配取最长者，但顺位更易读且不易误删）
    static = cfg.find("alias /srv/hddiv/site/")
    assert static == -1 or proxy < static, "API 反代块应排在静态规则之前"
    # location 与 proxy_pass 的末尾斜杠必须成对，否则路径会被改写错
    assert re.search(r"location\s+/ggx/api/\s*\{", cfg)
    assert "proxy_pass         http://127.0.0.1:8099/api/;" in cfg


def test_launchd_template_is_valid_and_placeholder_based() -> None:
    """launchd 模板必须可渲染：占位符齐全、plist 结构完整。"""
    import plistlib

    tpl = (project_root() / "deploy" / "com.hddiv.web.plist.example")
    assert tpl.is_file(), "缺少 launchd 模板"
    raw = tpl.read_text(encoding="utf-8")
    for ph in ("__PYTHON__", "__PROJECT_ROOT__"):
        assert ph in raw, f"模板缺少占位符 {ph}"
    # 注释里也有 __，解析时先剥掉注释
    body = re.sub(r"<!--.*?-->", "", raw, flags=re.DOTALL)
    data = plistlib.loads(body.encode("utf-8"))
    assert data["Label"] == "com.hddiv.web"
    assert data["RunAtLoad"] is True
    assert data["KeepAlive"] == {"SuccessfulExit": False}, "应支持崩溃自愈"
    assert "--api-only" in data["ProgramArguments"]


def test_install_service_script_placeholder_guard() -> None:
    """安装脚本必须检测未替换的占位符 —— 否则 launchd 会静默失败。"""
    sh = (project_root() / "deploy" / "install-service.sh").read_text(encoding="utf-8")
    assert "plutil -lint" in sh, "应校验 plist 格式"
    assert "未替换的占位符" in sh or "grep -q \"__\"" in sh
    assert "launchctl load" in sh and "launchctl unload" in sh


def test_serve_script_is_syntax_valid() -> None:
    import subprocess

    sh = project_root() / "deploy" / "serve.sh"
    r = subprocess.run(["bash", "-n", str(sh)], capture_output=True, text=True)
    assert r.returncode == 0, f"serve.sh 语法错误：{r.stderr}"


@requires_db
def test_members_default_returns_selected_only() -> None:
    """回归：不传 passed 时应返回「入选」股票，而不是全部候选。

    曾默认返回全部候选（5000+ 只），与「股票清单」的语义不符。
    """
    from hdiv.web import service

    runs = service.list_universes()
    if not runs:
        pytest.skip("没有筛选记录")
    rid = runs[0]["run_id"]
    default = service.list_members(rid, passed=True, size=1)
    allc = service.list_members(rid, passed=None, size=1)
    rejected = service.list_members(rid, passed=False, size=1)
    assert default["total"] <= allc["total"]
    assert default["total"] + rejected["total"] == allc["total"], \
        "入选数 + 淘汰数 应等于候选总数"
    assert default["total"] > 0, "示例记录应有入选股票"


# ---------------------------------------------------------------------------
# HTML 报告降级为「导出件」
# ---------------------------------------------------------------------------


def test_html_flag_is_opt_in_in_cli() -> None:
    """回归：静态报告已降级为导出件 —— 默认不生成，需显式 --html。

    改造后 SPA 是主界面，默认再产出 HTML 会：
      1) 每次筛选多一个文件
      2) 同 asof 重跑时按 asof 命名互相覆盖，与「每次运行都留痕」矛盾
    """
    from hdiv.cli import build_parser

    p = build_parser()
    subs = {a.dest: a for a in p._actions if hasattr(a, "choices") and isinstance(a.choices, dict)}
    sub = subs["command"].choices
    for name in ("universe", "profile", "backtest", "audit", "sensitivity"):
        args = p.parse_args([name] if name != "universe" else [name])
        assert getattr(args, "html") is False, f"hdiv {name} 的 --html 应为 opt-in"
        assert hasattr(args, "no_html"), f"hdiv {name} 应保留 --no-html 以免旧命令报错"
        _ = sub  # 断言子命令存在


def test_default_universe_run_writes_no_html(monkeypatch) -> None:
    """不传 --html 时不应调用 HTML 生成器。"""
    from hdiv import cli

    called = {"n": 0}
    import hdiv.report.build as build

    def fake(*a, **k):
        called["n"] += 1
        raise AssertionError("默认不应生成 HTML")

    monkeypatch.setattr(build, "build_universe_report", fake)
    # 仅验证解析结果：默认 html=False
    p = cli.build_parser()
    assert p.parse_args(["universe"]).html is False
    assert p.parse_args(["universe", "--html"]).html is True
    assert called["n"] == 0


def test_report_names_include_run_id() -> None:
    """报告文件名必须带执行 id，否则同 asof/同 symbol 重跑会互相覆盖。"""
    from hdiv.core.config import load_config

    n = load_config("report").naming
    for key in ("universe", "profile", "backtest", "walkforward", "sensitivity"):
        pattern = getattr(n, key)
        assert "{run_id}" in pattern or "{wf_id}" in pattern or "{sens_id}" in pattern, \
            f"naming.{key} 缺少执行 id 占位符，重跑会覆盖：{pattern}"


@requires_db
def test_universe_report_filenames_do_not_collide() -> None:
    """同 asof 的两条记录必须产出两个不同文件（实测回归）。"""
    from hdiv.core.config import load_config
    from hdiv.report.renderer import Renderer

    r = Renderer()
    a = r.name_from("universe", asof="2025-01-21", run_id="a" * 32)
    b = r.name_from("universe", asof="2025-01-21", run_id="b" * 32)
    assert a != b, "同 asof 不同 run 必须产生不同文件名"
    assert load_config("report").naming.universe.startswith("reports/")


@requires_db
def test_universe_detail_exposes_chart_data() -> None:
    """前端图表所需数据由后端算好（前端不做业务计算）。"""
    from hdiv.web import service

    runs = service.list_universes()
    if not runs:
        pytest.skip("没有筛选记录")
    u = service.get_universe(runs[0]["run_id"])
    f = u["funnel"]
    assert len(f["labels"]) == len(f["values"]) == 5
    # 漏斗最后一段必须等于入选数，否则 stats 不自洽
    assert f["values"][-1] == u["member_count"]
    assert f["values"][0] == u["candidate_count"]
    # 存活数必须单调不增
    assert all(f["values"][i] >= f["values"][i + 1] for i in range(len(f["values"]) - 1)), \
        f"漏斗存活数应单调不增：{f['values']}"
    dist = u["industry_distribution"]
    assert isinstance(dist, list)
    if dist:
        assert {"industry", "count"} <= set(dist[0])
        assert dist[0]["count"] >= dist[-1]["count"], "行业分布应按数量降序"


# ---------------------------------------------------------------------------
# 站点根 index.html 的所有权（曾发生真实事故）
# ---------------------------------------------------------------------------


def test_naming_index_is_under_reports() -> None:
    """回归：静态报告索引曾与前端首页抢 output/index.html。

    命名配置里 index 漏配 reports/ 前缀 + 调用点硬编码 "index.html"，
    导致跑一次 `hdiv audit --html`（内部会 build_index）就把前端首页
    覆盖成静态报告索引。
    """
    from hdiv.core.config import load_config

    naming = load_config("report").naming
    assert naming.index == "reports/index.html", \
        f"naming.index 必须是 reports/index.html（当前 {naming.index}）"
    # 所有静态报告都应在 reports/ 下，不占用站点根
    for key in ("index", "audit", "universe", "profile", "backtest",
                "walkforward", "sensitivity"):
        pattern = getattr(naming, key)
        assert pattern.startswith("reports/"), \
            f"naming.{key} 应输出到 reports/ 下，不占用站点根：{pattern}"


def test_renderer_refuses_to_write_site_root_index() -> None:
    """渲染器必须拒绝把报告写到站点根 index.html（前置拦截）。"""
    from hdiv.core.errors import HdivError
    from hdiv.report.renderer import Renderer

    r = Renderer()
    for bad in ("index.html", "./index.html"):
        with pytest.raises(HdivError, match="统一前端首页冲突"):
            r.render("reports/index.html", {}, bad, report_type="index")


def test_site_root_index_is_spa() -> None:
    """站点根 index.html 必须是前端外壳（引用 app/app.js），不是静态报告索引。"""
    idx = project_root() / "output" / "index.html"
    if not idx.is_file():
        pytest.skip("尚未生成站点")
    html = idx.read_text(encoding="utf-8")
    assert "app/app.js" in html, "output/index.html 不是统一前端外壳（可能被报告覆盖）"
    assert "报告索引" not in html or "app/app.js" in html


def test_build_index_uses_naming_config() -> None:
    """build_index 的输出路径必须来自 naming 配置，不能硬编码。"""
    import inspect

    from hdiv.report import build

    src = inspect.getsource(build.build_index)
    assert 'r.name_from("index")' in src, "build_index 应使用 naming 配置生成文件名"
    # 不应再把 "index.html" 当输出名直接传进去
    assert '        "index.html",\n' not in src, "build_index 仍在硬编码输出名"


def test_site_build_does_not_clobber_spa() -> None:
    """site.sync_frontend 之后，站点根 index.html 必须仍是 SPA。"""
    from hdiv.web import site

    site.sync_frontend(verbose=False)
    html = (project_root() / "output" / "index.html").read_text(encoding="utf-8")
    assert "app/app.js" in html


# ---------------------------------------------------------------------------
# 回测内分析：任意日持仓 + 个股买卖点
# ---------------------------------------------------------------------------


def _sample_backtest_run() -> str | None:
    from hdiv.core.config import load_config
    from hdiv.data import db

    df = db.read_sql(
        "SELECT r.run_id FROM hd_backtest_run r "
        "JOIN hd_backtest_position p ON p.run_id = r.run_id "
        "WHERE r.mode = 'single' "
        "GROUP BY r.run_id ORDER BY COUNT(*) DESC LIMIT 1",
        cfg=load_config("datasource"),
    )
    return None if df.empty else str(df["run_id"].iloc[0])


@requires_db
def test_position_dates_is_compact_by_default() -> None:
    """默认只返回日期字符串：带全字段会让响应从约 30KB 涨到 460KB。"""
    from hdiv.web import analysis

    rid = _sample_backtest_run()
    if not rid:
        pytest.skip("没有带持仓的回测")
    d = analysis.position_dates(rid)
    assert "dates" in d and d["dates"], "应返回日期数组"
    assert "items" not in d, "默认不应返回逐日全字段明细"
    assert d["count"] == len(d["dates"])
    assert d["dates"] == sorted(d["dates"]), "日期应升序"
    # 紧凑形式必须显著更小
    import json

    compact = len(json.dumps(d, ensure_ascii=False).encode())
    full = len(json.dumps(analysis.position_dates(rid, detail=True),
                          ensure_ascii=False).encode())
    assert compact < full / 3, f"紧凑形式应远小于明细（{compact} vs {full}）"


@requires_db
def test_portfolio_falls_back_to_previous_trading_day() -> None:
    """非交易日应回退到之前最近的有快照交易日，并如实标注。"""
    from hdiv.web import analysis
    from hdiv.core.errors import HdivError

    rid = _sample_backtest_run()
    if not rid:
        pytest.skip("没有带持仓的回测")
    dates = analysis.position_dates(rid)["dates"]
    d = analysis.portfolio_on_date(rid, dates[-1])
    assert d["date"] == dates[-1] and not d["adjusted"]

    # 区间内但非交易日（用周末构造）
    import datetime as _dt

    mid = _dt.date.fromisoformat(dates[len(dates) // 2])
    weekend = mid + _dt.timedelta(days=(5 - mid.weekday()) % 7 + 1)
    d2 = analysis.portfolio_on_date(rid, weekend.isoformat())
    assert d2["date"] <= weekend.isoformat()
    assert d2["adjusted"] is True, "非交易日应标注已回退"

    # 早于首个快照应给出可理解错误
    with pytest.raises(HdivError, match="早于该回测的首个快照"):
        analysis.portfolio_on_date(rid, "1990-01-01")


@requires_db
def test_portfolio_summary_is_internally_consistent() -> None:
    """持仓汇总必须自洽：市值合计 = 逐股之和；权重合计 ≈ 仓位占比。"""
    from hdiv.web import analysis

    rid = _sample_backtest_run()
    if not rid:
        pytest.skip("没有带持仓的回测")
    dates = analysis.position_dates(rid)["dates"]
    # 找一个有持仓的交易日
    for day in reversed(dates):
        d = analysis.portfolio_on_date(rid, day)
        if d["positions"]:
            break
    else:
        pytest.skip("没有非空持仓日")

    s = sum(p["market_value"] or 0 for p in d["positions"])
    assert abs(s - d["summary"]["market_value"]) < 1.0
    assert d["summary"]["count"] == len(d["positions"])
    w = sum(p["weight"] or 0 for p in d["positions"])
    tv = d["equity"]["total_value"] or 0
    if tv:
        assert abs(w - d["summary"]["market_value"] / tv) < 0.02, \
            f"权重合计 {w:.4f} 应约等于仓位占比 {d['summary']['market_value']/tv:.4f}"
    # 每只股票都应带名称（JOIN stock）
    assert all(p["symbol"] for p in d["positions"])


@requires_db
def test_stock_detail_series_and_trades() -> None:
    """个股买卖点：序列长度一致、买卖点带完整成交信息。"""
    from hdiv.web import analysis

    rid = _sample_backtest_run()
    if not rid:
        pytest.skip("没有带持仓的回测")
    stocks = analysis.run_stocks(rid)
    if not stocks:
        pytest.skip("该回测没有持仓股票")
    sym = stocks[0]["symbol"]

    d = analysis.stock_detail(rid, sym)
    n = len(d["dates"])
    assert n > 0
    for k, v in d["series"].items():
        assert len(v) == n, f"序列 {k} 长度与日期不一致（{len(v)} vs {n}）"
    assert "close" in d["series"]

    # 股息率必须在合理量级内（单位错误会让它变成 0 或几百）
    dv = [x for x in d["series"]["dv_yield"] if x is not None]
    if dv:
        assert max(dv) < 1.0, f"股息率不应超过 100%：{max(dv)}"
        assert min(dv) >= 0.0, "股息率不应为负"

    for t in d["trades"]:
        assert t["side"] in {"BUY", "SELL"}
        assert t["price"] and t["price"] > 0
        assert t["quantity"] and t["quantity"] > 0
        assert t["amount"] and t["amount"] > 0
        assert t["reason_text"] and t["reason_text"] != ""
    assert d["stats"]["trade_count"] == len(d["trades"])


@requires_db
def test_stock_detail_with_sell_trades_does_not_500() -> None:
    """回归：有卖出的个股必须能打开。

    卖出成交的 ``holding_days`` 在库里是 NaN 而**不是** None，
    老代码 ``int(r["holding_days"]) if ... is not None else None`` 会抛
    ``ValueError: cannot convert float NaN to integer``，
    让个股详情接口 500 —— 22/35 只有卖出的个股整页打不开。
    """
    from hdiv.web import analysis

    rid = _sample_backtest_run()
    if not rid:
        pytest.skip("没有带持仓的回测")
    stocks = analysis.run_stocks(rid)
    sells = [s for s in stocks if (s.get("sell_count") or 0) > 0]
    if not sells:
        pytest.skip("该回测没有卖出成交")
    sym = sells[0]["symbol"]

    d = analysis.stock_detail(rid, sym)      # 老代码在这一行 500
    assert any(t["side"] == "SELL" for t in d["trades"]), "应至少有一笔卖出"
    for t in d["trades"]:
        hd = t["holding_days"]
        assert hd is None or isinstance(hd, int), f"holding_days 应为整数或 None：{hd!r}"
        assert hd is None or hd >= 0
    # NaN 会以非法 JSON 的形式漏到前端，这里一并卡住
    json.dumps(d, allow_nan=False)


@requires_db
def test_stock_detail_respects_series_selection() -> None:
    """勾选哪些指标就只算哪些（不为没勾的做无谓计算）。"""
    from hdiv.web import analysis
    from hdiv.core.errors import HdivError

    rid = _sample_backtest_run()
    if not rid:
        pytest.skip("没有带持仓的回测")
    sym = analysis.run_stocks(rid)[0]["symbol"]
    d = analysis.stock_detail(rid, sym, series=["close", "roe"])
    assert set(d["series"]) == {"close", "roe"}
    assert "pe_ttm" not in d["series"]
    # 无效指标应报错而不是静默忽略
    with pytest.raises(HdivError):
        analysis.stock_detail(rid, sym, series=["不存在的指标"])


@requires_db
def test_roe_series_is_stepwise_not_interpolated() -> None:
    """ROE 必须按公告日对齐成阶梯（插值会造出当时不存在的值）。"""
    from hdiv.web import analysis

    rid = _sample_backtest_run()
    if not rid:
        pytest.skip("没有带持仓的回测")
    sym = analysis.run_stocks(rid)[0]["symbol"]
    d = analysis.stock_detail(rid, sym, series=["roe"])
    vals = [x for x in d["series"]["roe"] if x is not None]
    if len(vals) < 50:
        pytest.skip("ROE 样本不足")
    # 阶梯序列的不同取值数应远少于样本数（季度更新，约 4 次/年）
    distinct = len(set(round(v, 6) for v in vals))
    assert distinct < len(vals) / 5, \
        f"ROE 取值数 {distinct} 相对样本 {len(vals)} 过多，疑似插值而非阶梯"


# ---------------------------------------------------------------------------
# Walk-forward 前端可见性
# ---------------------------------------------------------------------------


@requires_db
def test_walkforward_list_and_detail() -> None:
    """Walk-forward 记录必须在接口层可见（此前完全没有入口）。"""
    from hdiv.web import service

    items = service.list_walkforwards()
    if not items:
        pytest.skip("没有 walk-forward 记录")
    w = items[0]
    assert w["wf_id"] and w["window_count"] > 0
    assert w["title"], "应有可读标题"
    assert w["strategy"]["conditions"], "应带策略条件说明"
    assert "oos" in w and w["oos"].get("window_count") == w["window_count"]

    d = service.get_walkforward(w["wf_id"])
    assert d is not None
    assert len(d["windows"]) == w["window_count"]
    for win in d["windows"]:
        # 每个窗口都必须有训练段与测试段
        assert win["train_start"] and win["test_start"]
        assert win["train_run_id"] and win["test_run_id"]
        assert "in_sample" in win and "out_of_sample" in win
    s = d["summary"]
    assert len(s["oos_returns"]) == w["window_count"]
    assert s["oos_mean"] is not None
    assert 0.0 <= s["oos_win_rate"] <= 1.0


@requires_db
def test_walkforward_summary_is_consistent() -> None:
    """汇总必须与逐窗口数据自洽（曾靠 metric 行数反推导致胜率算错）。"""
    from hdiv.web import service

    items = service.list_walkforwards()
    if not items:
        pytest.skip("没有 walk-forward 记录")
    for w in items[:3]:
        d = service.get_walkforward(w["wf_id"])
        rets = [x["out_of_sample"].get("total_return") for x in d["windows"]]
        rets = [x for x in rets if x is not None]
        if not rets:
            continue
        assert abs(d["summary"]["oos_mean"] - sum(rets) / len(rets)) < 1e-9
        expect_win = sum(1 for x in rets if x > 0) / len(rets)
        assert abs(d["summary"]["oos_win_rate"] - expect_win) < 1e-9
        # 列表页的汇总应与详情页一致
        assert abs((w["oos"]["mean_return"] or 0) - d["summary"]["oos_mean"]) < 1e-9


def test_walkforward_frontend_page_exists() -> None:
    """前端必须有 walk-forward 页与导航入口。"""
    js = (project_root() / "web" / "app.js").read_text(encoding="utf-8")
    html = (project_root() / "web" / "index.html").read_text(encoding="utf-8")
    assert "viewWalkforwards" in js and "viewWalkforwardDetail" in js
    assert "#/walkforwards" in html, "导航缺「样本外」入口"
    assert "mountWalkforwardDetail" in js, "详情页应挂载对比图"


@requires_db
def test_walkforward_exposes_benchmark_and_excess() -> None:
    """回归：基准指标存为 benchmark_<code>，曾被 benchmark_code='' 过滤掉，
    导致页面上看不到最重要的「超额收益」。"""
    from hdiv.web import service

    items = service.list_walkforwards()
    if not items:
        pytest.skip("没有 walk-forward 记录")
    d = service.get_walkforward(items[0]["wf_id"])
    s = d["summary"]
    assert s.get("benchmark_mean") is not None, "缺少基准均值"
    assert s.get("excess_mean") is not None, "缺少超额收益均值"
    assert s.get("excess_win_rate") is not None

    got = 0
    for w in d["windows"]:
        if w["benchmark_return"] is None:
            continue
        got += 1
        assert w["benchmark_code"], "应记录基准代码"
        o = w["out_of_sample"].get("total_return")
        if o is not None:
            assert abs(w["excess_return"] - (o - w["benchmark_return"])) < 1e-9, \
                "超额必须等于 策略收益 − 基准收益"
        # 基准行不得混进策略指标里
        assert not any(k.startswith("benchmark::") for k in w["out_of_sample"])
    assert got > 0, "没有任何窗口带基准收益"


@requires_db
def test_walkforward_frozen_params_record_calibration() -> None:
    """冻结参数必须记录「校准出的绝对阈值」，而不只是配置里的分位。

    训练段的作用是把相对分位（P75）转成绝对股息率；若只有分位、
    没有绝对阈值，说明训练段实际上没做校准。
    """
    from hdiv.web import service

    items = service.list_walkforwards()
    if not items:
        pytest.skip("没有 walk-forward 记录")
    d = service.get_walkforward(items[0]["wf_id"])
    abs_entries = []
    for w in d["windows"]:
        f = w["frozen_params"]
        assert "entry_yield_percentile" in f, "应保留分位口径"
        assert f.get("absolute_entry_yield"), f"窗口 {w['window_index']} 缺校准阈值"
        assert f.get("calibration_obs"), "应记录校准样本数"
        abs_entries.append(f["absolute_entry_yield"])
    # 各窗口的绝对阈值应随市场水平变化（全相同说明没真校准）
    assert len(set(round(x, 6) for x in abs_entries)) > 1, \
        "各窗口校准出的绝对阈值完全相同，疑似未真正校准"