# -*- coding: utf-8 -*-
"""育种统计 Schema（一期 2 个最基础：ABLUP·EBV + 配合力 GCA/SCA）"""
from datetime import datetime

from pydantic import BaseModel, ConfigDict, Field


class RunStatsIn(BaseModel):
    """提交统计任务的入参。"""
    trait_id: int = Field(..., description="性状 id（bre_trait）")
    trait_code: str = Field(..., description="性状编码（bre_trait.trait_code，如 avg_fruit_weight）")
    year: int | None = Field(None, description="评价年份过滤；为空则全部")
    fixed_effects: list[str] | None = Field(
        None, description="固定效应因子：支持 trial_study(试验/站点) / rootstock(砧木)"
    )
    covariate: str | None = Field(
        None, description="协变量（BLUP 校正）：crop_load(负载量) / competition(空间竞争:相邻株数)"
    )
    data_gate: bool = Field(
        True, description="数据就绪门禁(G1)：每 clone/家系最小 n、系谱完整率、缺失率任一不达标则拒绝运行"
    )
    min_clone_n: int = Field(
        2, ge=0, description="每 clone/家系最小样本量（观测树数低于该值的组列入未达标）"
    )
    min_pedigree_rate: float = Field(
        0.3, ge=0.0, le=1.0, description="系谱完整率下限（有父母或组合亲本的观测树占比）"
    )
    max_missing_rate: float = Field(
        0.8, ge=0.0, le=1.0, description="表型缺失率上限（缺失=有评价记录但无该性状数值）"
    )
    gxe: bool = Field(
        False, description="启用 G×E 交互随机效应（clone×site / 组合×site；year 同理）"
    )
    gxe_env: str = Field(
        "site", description="G×E 环境维度：site(trial_study_id) / year(evaluate_year)"
    )
    gxr: bool = Field(
        False, description="启用砧木×接穗随机互作（G×R）：rootstock 从固定效应移到随机效应，与 gxe 互斥"
    )
    spatial: bool = Field(
        False,
        description=(
            "启用 AR1×AR1 空间协方差（残差 e~N(0,σ²e·R)，R_ij=ρ^(|Δrow|+|Δcol|)）："
            "需观测株均有 bre_tree.row_no/col_no；与 gxe/gxr 互斥"
        ),
    )
    spatial_aniso: bool = Field(
        False,
        description=(
            "空间协方差各向异性双参数（需 spatial=True）：R_ij=ρ_row^|Δrow|·ρ_col^|Δcol|，"
            "方法 AR1×AR1(aniso)，输出 ρ_row/ρ_col；False 时用单参 ρ（v1 兼容）"
        ),
    )
    block: bool = Field(
        False,
        description=(
            "启用区组随机效应（不完全区组/增广/α-格子）：bre_tree.block_no 作第二随机效应，"
            "精度收益进入遗传评估；需 ≥2 个区组；与 gxe/gxr/spatial 互斥"
        ),
    )
    stage: str | None = Field(
        None, description="发育阶段拆分：juvenile(童期)/evaluation(成株)；给定则只取该阶段观测建模并拆独立批次"
    )
    design_type: str = Field(
        "full_diallel",
        description=(
            "交配设计（配合力分析用）：full_diallel=完全双列(Griffing) / "
            "partial_diallel=部分双列 / line_tester=line×tester(NCII 两因素模型) / "
            "nciii=NCIII 测交"
        ),
    )
    model_config = ConfigDict(extra="ignore")


class PredictionOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    model_name: str | None = None
    trait_id: int | None = None
    method: str | None = None
    stage: str | None = None
    accuracy: float | None = None
    train_n: int | None = None
    heritability: float | None = None
    data_version: str | None = None
    input_hash: str | None = None
    engine_version: str | None = None
    is_active: bool | None = None
    note: str | None = None
    created_time: datetime | None = None


class PredictionValueOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    prediction_id: int
    germplasm_id: int | None = None
    tree_id: int | None = None
    trait_id: int | None = None
    predicted_value: float | None = None
    reliability: float | None = None
    pa: float | None = None
    rank: int | None = None


class CombiningAbilityOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    model_name: str | None = None
    trait_id: int | None = None
    method: str | None = None
    design_type: str | None = None
    gca_json: dict | None = None
    sca_json: dict | None = None
    anova_json: dict | None = None


class StatisticsJobOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    job_type: str | None = None
    status: str
    params_json: dict | None = None
    result_ref: int | None = None
    error_msg: str | None = None
    started_time: datetime | None = None
    finished_time: datetime | None = None


# ---------- 纯 Python 统计分析（不依赖 R） ----------
class DescribeStatsIn(BaseModel):
    """描述性统计入参。"""
    trait_codes: list[str] | None = Field(None, description="数值性状列名；为空则用全部数值列")
    group_by: str = Field("none", description="分组维度：none/year/combination")
    year: int | None = Field(None, description="评价年份过滤")
    model_config = ConfigDict(extra="ignore")


class CorrelationIn(BaseModel):
    """相关分析入参。"""
    trait_codes: list[str] | None = Field(None, description="参与相关的数值性状；至少2个")
    mode: str = Field(
        "pheno",
        description="相关类型：pheno=表型相关（皮尔逊，纯Python）；genetic=遗传相关（成对双性状BLUP REML）",
    )
    g_method: str = Field(
        "mtblup",
        description="遗传相关 r_g 来源（仅 mode=genetic）：mtblup=成对双性状BLUP(REML精确)；calo=可靠性校正EBV相关，无生效EBV批次时回退mtblup",
    )
    year: int | None = Field(None, description="评价年份过滤")
    model_config = ConfigDict(extra="ignore")


class SelectionIndexIn(BaseModel):
    """选择指数入参。"""
    weights: dict[str, float] | None = Field(None, description="性状->权重，缺省等权；自动归一化")
    year: int | None = Field(None, description="评价年份过滤（仅表型回退路径使用）")
    top_n: int = Field(50, description="返回排名前 N 的单株")
    batch_ids: dict[str, int] | None = Field(None, description="性状->ABLUP批次id，指定则用该批次的EBV作为指数输入")
    use_h2: bool = Field(True, description="有效权重 = 用户权重 × 遗传力(h²) 加权（仅 zsum 模式生效）")
    method: str = Field(
        "zsum",
        description=(
            "zsum=加权Z综合（轻量，可表型回退）；"
            "smith_hazel=真Smith-Hazel b=P⁻¹Ga（需各性状指定EBV批次且含h²）；"
            "restricted=约束指数 b=P⁻¹G(I−M)a（Kempthorne-Nordskog，受限性状 ΔG=0，需 restricted_traits）"
        ),
    )
    aggregate: str = Field(
        "clone",
        description=(
            "聚合粒度：clone=无性系级（同系多株可靠加权聚合为一个遗传实体再排名，桃无性繁殖默认）；"
            "tree=单株级（旧行为）"
        ),
    )
    stage: str | None = Field(
        None, description="发育阶段过滤：juvenile(童期)/evaluation(成株)；不传且性状横跨两阶段时返回 stage_warning"
    )
    g_method: str = Field(
        "calo", description="遗传相关 r_g 来源（smith_hazel / restricted 生效）：calo=可靠性校正EBV相关；mtblup=成对双性状BLUP(REML精确估计，单对不收敛回退calo)"
    )
    auto_weights: bool = Field(
        False,
        description="自动权重：以各性状 default_h2（实测h²优先，兜底0.1）为权重；强制 use_h2=False 防双重相乘",
    )
    restricted_traits: list[str] | None = Field(
        None, description="受限性状列表（仅 method=restricted 生效）：这些性状的遗传增益被约束为 0，其余性状自由响应（Kempthorne-Nordskog 闭式投影）"
    )
    economic_weights: dict[str, float] | None = Field(
        None, description="经济权重（经济价值，绝对尺度不归一化；smith_hazel / restricted 生效）：直接作为聚合基因型系数 a，未列出性状按 0 处理"
    )
    model_config = ConfigDict(extra="ignore")


class SelectionIndexOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    model_name: str | None = None
    method: str | None = None
    weights_json: dict | None = None
    batch_refs_json: dict | None = None
    heritability_json: dict | None = None
    top_n: int | None = None
    result_json: dict | None = None
    created_time: datetime | None = None


class SelectionIndexApplyIn(BaseModel):
    """选择指数 -> 决选 入参。"""
    top_n: int = Field(50, description="写入前 N 名单株")
    selection_year: int | None = Field(None, description="入选年份；为空用当年")
    rule_id: int | None = Field(None, description="来源选择规则(决策预览选定,用于溯源 rule_id/from_stage/to_stage)")
    min_reliability: float = Field(
        0.2, ge=0.0, le=1.0,
        description="EBV 可靠性硬门槛：指数引用批次的 EBV 可靠性低于该值的单株跳过(证据不足不入选)",
    )
    model_config = ConfigDict(extra="ignore")


class KinshipIn(BaseModel):
    """亲缘/近交分析入参。"""
    threshold: float = Field(0.25, description="亲缘系数预警阈值（r>A 预警）")
    tree_ids: list[int] | None = Field(None, description="限定单株范围；为空用全部活动单株")
    model_config = ConfigDict(extra="ignore")


class DataQualityIn(BaseModel):
    """数据质量/异常值诊断入参。"""
    trait_id: int = Field(..., description="性状 id（bre_trait）")
    trait_code: str = Field(..., description="性状编码")
    year: int | None = Field(None, description="评价年份过滤；为空则全部")
    trial_study_id: int | None = Field(None, description="限定试验/站点(MET)；为空则全部")
    dataset_id: int | None = Field(
        None, description="基因型数据集：提供时附加孟德尔检验段（最新批次计数+flag树+隔离状态）")
    model_config = ConfigDict(extra="ignore")


class GeneticGainIn(BaseModel):
    """ΔG 遗传增益投影入参。"""
    trait_id: int = Field(..., description="性状 id（bre_trait）")
    trait_code: str = Field(..., description="性状编码")
    prediction_id: int | None = Field(None, description="指定 EBV 批次；为空用该性状最新批次")
    top_p: float | None = Field(None, gt=0, lt=1, description="入选比例（0-1，与 top_n 二选一）")
    top_n: int | None = Field(None, ge=1, description="入选株数（与 top_p 二选一）")
    generation_interval: float = Field(1.0, gt=0, description="世代间隔(年)；年增益=ΔG/L")
    model_config = ConfigDict(extra="ignore")


class InbreedingDepressionIn(BaseModel):
    """近交衰退分析入参。"""
    trait_id: int = Field(..., description="性状 id（bre_trait）")
    trait_code: str = Field(..., description="性状编码")
    year: int | None = Field(None, description="评价年份过滤；为空则全部")
    trial_study_id: int | None = Field(None, description="限定试验/站点(MET)；为空则全部")
    min_n: int = Field(10, ge=2, description="回归最小样本量（同时具备 F 与表型的树）")
    model_config = ConfigDict(extra="ignore")


class TrialDesignIn(BaseModel):
    """试验设计生成入参（rcbd / augmented / alpha）。"""
    trial_study_id: int = Field(..., description="试验研究点 id（bre_trial_study）")
    design_type: str = Field("rcbd", description="设计类型：rcbd=随机完全区组、augmented=增广、alpha=α-格子")
    seed: int | None = Field(None, description="随机种子；固定 seed 可复现同一设计")
    check_germplasm_ids: list[int] = Field(default_factory=list,
                                           description="增广设计对照种质 id 列表（每区组重复）")
    block_size: int | None = Field(None, description="α-格子区组大小 k（每重复区组块大小）")
    reps: int | None = Field(None, description="α-格子重复数 r（平方格子缺省 k+1）")
    model_config = ConfigDict(extra="ignore")


class MatingRecommendIn(BaseModel):
    """主动选配推荐入参。"""
    candidate_germplasm_ids: list[int] = Field(..., description="候选亲本种质 id 集合（≥2）")
    prediction_id: int | None = Field(None, description="EBV 批次；为空用最新 ABLUP/GBLUP 批次")
    kinship_threshold: float = Field(0.25, description="亲缘惩罚阈值（r>阈值开始扣分）")
    w_ebv: float = Field(1.0, ge=0, description="EBV 互补权重")
    w_kin: float = Field(1.0, ge=0, description="近交惩罚权重")
    max_pairs: int = Field(20, ge=1, le=500, description="返回 top N 配对")
    model_config = ConfigDict(extra="ignore")


class OcsIn(BaseModel):
    """最优贡献选择（OCS）入参。"""
    candidate_germplasm_ids: list[int] = Field(..., description="候选亲本种质 id 集合")
    n_select: int = Field(5, ge=1, description="选择数量（贡献总和 Σc=n_select）")
    lam: float = Field(0.1, ge=0, description="近交约束权重 λ：越大越压低子代平均亲缘（avg_kinship 单调非增），0=无近交约束（满额投最高 EBV 单亲）")
    prediction_id: int | None = Field(None, description="EBV 批次；为空用最新 ABLUP/GBLUP 批次")
    model_config = ConfigDict(extra="ignore")


class MabcIn(BaseModel):
    """MABC 标记辅助回交进度入参。"""
    candidate_tree_ids: list[int] = Field(..., description="候选回交单株树 id 集合（BC 世代分离群体）")
    foreground_panel_ids: list[int] = Field(..., description="前景选择面板 id（目标性状 MAS 面板）")
    background_panel_ids: list[int] = Field(..., description="背景恢复面板 id（全基因组标记面板）")
    recurrent_parent_tree_id: int = Field(..., description="轮回亲本树 id（背景纯合一致对照）")
    foreground_min_hits: int = Field(1, ge=1, description="前景通过阈值（有利剂量命中标记数 ≥ 此值，stage 无关、童期可用）")
    generation: str = Field("BC1", description="候选当前回交代（F1/BC1/BC2/BC3），用于晋级建议")
    background_target: float = Field(90.0, gt=0, le=100, description="背景恢复目标 %（达到则建议晋级下一回交代）")
    model_config = ConfigDict(extra="ignore")


class AnovaIn(BaseModel):
    """ANOVA / 广义遗传力 H² 入参。"""
    trait_id: int = Field(..., description="性状 id（bre_trait）")
    trait_code: str = Field(..., description="性状编码")
    year: int | None = Field(None, description="评价年份过滤；为空则全部")
    block: bool = Field(
        False, description="RCBD 设计基 ANOVA：True=从残差析出区组效应(SS_block/F_block/p_block)，遗传差异为区组校正后；需观测株 bre_tree.block_no")
    model_config = ConfigDict(extra="ignore")


class AnovaOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    model_name: str | None = None
    trait_id: int | None = None
    method: str | None = None
    result_json: dict | None = None
    created_time: datetime | None = None


class CvRunIn(BaseModel):
    """k-fold 交叉验证入参。"""
    trait_id: int = Field(..., description="性状 id（bre_trait）")
    trait_code: str = Field(..., description="性状编码")
    year: int | None = Field(None, description="评价年份过滤；为空则全部")
    fixed_effects: list[str] | None = Field(
        None, description="固定效应因子：trial_study(试验/站点) / rootstock(砧木)"
    )
    covariate: str | None = Field(
        None, description="协变量：crop_load(负载量) / competition(空间竞争)"
    )
    gxe: bool = Field(False, description="启用 G×E 交互随机效应")
    gxe_env: str = Field("site", description="G×E 环境维度：site / year")
    k: int = Field(5, ge=2, le=10, description="折数")
    dataset_id: int | None = Field(
        None, description="基因型数据集 id；给定则走 GS 交叉验证（GBLUP/ssGBLUP/rrBLUP/BayesB）"
    )
    method: str | None = Field(
        None, description="GS 方法：gblup / ssgblup / rrblup / bayesb（仅 dataset_id 给定时生效）"
    )
    maf_min: float = Field(0.05, description="MAF 下限（GS 交叉验证用）")
    split: str = Field(
        "random", description="GS 折划分：random=固定种子随机分层（默认）/ family=家系阻塞折（同父半同胞同折，防亲缘泄漏）"
    )
    seed: int = Field(20260805, description="BayesB 随机种子（折划分与 Gibbs 共用，固定可复现）")
    model_config = ConfigDict(extra="ignore")


class CvResultOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    trait_id: int | None = None
    trait_code: str | None = None
    method: str | None = None
    k: int | None = None
    n_total: int | None = None
    n_individuals: int | None = None
    mean_pearson: float | None = None
    mean_rmse: float | None = None
    pooled_pearson: float | None = None
    pooled_rmse: float | None = None
    cv_accuracy: float | None = None
    h2: float | None = None
    data_version: str | None = None
    input_hash: str | None = None
    engine_version: str | None = None
    note: str | None = None
    created_time: datetime | None = None


class CvFoldOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    cv_result_id: int
    fold_idx: int | None = None
    n_train: int | None = None
    n_test: int | None = None
    n_eval: int | None = None
    pearson: float | None = None
    rmse: float | None = None
    error: str | None = None


class DecisionPreviewIn(BaseModel):
    """选择规则 × 表型/EBV/标记(MAS) 决策预览入参。"""
    rule_ids: list[int] | None = Field(None, description="指定规则；为空用全部启用规则")
    prediction_id: int | None = Field(None, description="EBV 源批次；为空用最新 ABLUP 批次")
    year: int | None = Field(None, description="表型均值年份过滤")
    min_reliability: float = Field(
        0.2, ge=0.0, le=1.0,
        description="EBV 可靠性门槛：单株 EBV 可靠性低于该值视为证据不足，不参与晋级/淘汰判定；"
                    "ebv 条件内可配 min_reliability 逐性状覆盖",
    )
    stage: str | None = Field(
        None, description="发育阶段过滤：juvenile(童期)/evaluation(成株)；不传且规则引用性状横跨两阶段时返回 stage_warning"
    )
    model_config = ConfigDict(extra="ignore")


class DecisionPreviewOut(BaseModel):
    rule_id: int | None = None
    rule_name: str | None = None
    action: str | None = None
    stage: str | None = None
    matched: bool | None = None
    reasons: list[str] | None = None
    tree_id: int | None = None
    tree_no: str | None = None
    combination_id: int | None = None


# ---------- AMMI / Finlay-Wilkinson 稳定性（§8.18） ----------
class StabilityRunIn(BaseModel):
    """稳定性分析入参。"""
    trait_id: int = Field(..., description="性状 id（bre_trait）")
    trait_code: str = Field(..., description="性状编码")
    gxe_env: str = Field("site", description="环境维度：site(trial_study_id) / year(evaluate_year)")
    year: int | None = Field(None, description="评价年份过滤；为空则全部")
    methods: list[str] = Field(
        ["ammi", "finlay"], description="分析方法：ammi / finlay，可多选"
    )
    model_config = ConfigDict(extra="ignore")


class StabilityOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    trait_id: int | None = None
    trait_code: str | None = None
    env_dim: str | None = None
    methods: str | None = None
    detail_json: dict | None = None
    data_version: str | None = None
    input_hash: str | None = None
    engine_version: str | None = None
    note: str | None = None
    created_time: datetime | None = None


# ---------- 遗传相关矩阵（MT-BLUP，§8.18） ----------
class GeneticCorrIn(BaseModel):
    """遗传相关矩阵入参。"""
    trait_ids: list[int] = Field(..., description="参与遗传相关的性状 id（≥2）")
    year: int | None = Field(None, description="评价年份过滤；为空则全部")
    full_mtblup: bool = Field(
        False, description="True=全多变量 EM-REML（一次估计完整 G0⊗A，替代逐对 bivariate）")
    model_config = ConfigDict(extra="ignore")


class GeneticCorrOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    trait_ids_json: dict | None = None
    matrix_json: dict | None = None
    sigma_a_json: dict | None = None
    heritability_json: dict | None = None
    pairs_json: dict | None = None
    n_common: int | None = None
    data_version: str | None = None
    input_hash: str | None = None
    engine_version: str | None = None
    note: str | None = None
    created_time: datetime | None = None


# ---------- Type-B 多环境遗传相关（§8.27） ----------
class TypeBIn(BaseModel):
    """Type-B 多环境遗传相关入参。"""
    trait_id: int = Field(..., description="性状 id（bre_trait，数值型）")
    trait_code: str = Field(..., description="性状编码")
    env_dim: str = Field("site", description="环境维度：site(研究点 trial_study_id) / year(评价年份) / stage(发育阶段：juvenile童期 vs evaluation成株，估幼年-成年遗传相关)")
    method: str = Field(
        "reml", description="方法：reml=环境互为性状逐对REML双性状BLUP(默认)；calo=分环境EBV相关/√rel(无环境级EBV时回退reml)"
    )
    year: int | None = Field(None, description="评价年份过滤；为空则全部")
    model_config = ConfigDict(extra="ignore")


class TypeBOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    trait_id: int | None = None
    trait_code: str | None = None
    env_dim: str | None = None
    method: str | None = None
    envs_json: dict | None = None
    matrix_json: dict | None = None
    pairs_json: dict | None = None
    n_common: int | None = None
    data_version: str | None = None
    input_hash: str | None = None
    engine_version: str | None = None
    note: str | None = None
    created_time: datetime | None = None


# ---------- UPGMA 聚类（§8.27，计算端点不落库） ----------
class ClusterIn(BaseModel):
    """UPGMA 聚类入参（实体=单株或无性系，按多性状轮廓聚类）。"""
    trait_ids: list[int] = Field(..., description="参与聚类的数值性状 id（≥2）")
    entity_type: str = Field("tree", description="聚类实体：tree=单株；clone=无性系(同系多株均值聚合)")
    mode: str = Field("pheno", description="值来源：pheno=表型均值；genetic=EBV(无生效批次回退表型)")
    distance: str = Field(
        "corr", description="实体距离：corr=1-|皮尔逊r|(轮廓相似)；euclidean=标准化欧氏距离"
    )
    k: int | None = Field(None, description="聚成 k 类；为空按合并距离最大跳变自动选择")
    model_config = ConfigDict(extra="ignore")


# ---------- GBLUP / ssGBLUP 基因组选择（§8.18） ----------
class GblupRunIn(BaseModel):
    """GBLUP/ssGBLUP/rrBLUP/BayesB 基因组选择入参（§8.18）。"""
    dataset_id: int = Field(..., description="基因型数据集 id（bre_genotyping_dataset）")
    trait_id: int = Field(..., description="性状 id（bre_trait）")
    trait_code: str = Field(..., description="性状编码")
    year: int | None = Field(None, description="评价年份过滤；为空则全部")
    method: str = Field("gblup", description="gblup（仅基因型株）/ ssgblup（单步法，含非基因型亲属）/ rrblup（岭回归逐标记，输出 marker effects）/ bayesb（贝叶斯可变选择，固定 seed 可复现）")
    maf_min: float = Field(0.05, ge=0.0, lt=0.5, description="MAF 过滤下限")
    seed: int = Field(20260805, description="BayesB 随机种子（固定保证可复现；MLOps 铁律）")
    model_config = ConfigDict(extra="ignore")


class GenotypingDatasetOut(BaseModel):
    id: int
    dataset_name: str | None = None
    platform: str | None = None
    panel: str | None = None
    purpose: str | None = None
    run_date: datetime | None = None
    n_samples: int | None = None


# ---------- GWAS / QTL / MAS 标记辅助选择（§8.22） ----------
class GwasRunIn(BaseModel):
    """GWAS 关联分析入参（GLM+PC 首版）。"""
    dataset_id: int = Field(..., description="基因型数据集 id（bre_genotyping_dataset）")
    trait_id: int = Field(..., description="性状 id（bre_trait）")
    trait_code: str = Field(..., description="性状编码")
    year: int | None = Field(None, description="评价年份过滤；为空则全部")
    method: str = Field("gwas", description="gwas（GLM+PC）/ emmax（混合模型 EMMAX）/ ssgwas（单步 GWAS）")
    maf_min: float = Field(0.05, ge=0.0, lt=0.5, description="MAF 过滤下限")
    n_pc: int = Field(3, ge=1, le=10, description="群体结构主成分数")
    sig_level: float = Field(0.05, gt=0.0, le=1.0, description="显著性水平")
    qtl_window: int = Field(1_000_000, ge=1, description="QTL 合并窗口(bp)")
    model_config = ConfigDict(extra="ignore")


class GwasQtlXEIn(BaseModel):
    """QTL×E 入参（按环境分层 GWAS + 稳定性判定，计算端点不建表）。"""
    dataset_id: int = Field(..., description="基因型数据集 id（bre_genotyping_dataset）")
    trait_id: int = Field(..., description="性状 id（bre_trait）")
    trait_code: str = Field(..., description="性状编码")
    year: int | None = Field(None, description="评价年份过滤；为空则全部")
    method: str = Field("gwas", description="gwas（GLM+PC）/ emmax（混合模型 EMMAX）")
    maf_min: float = Field(0.05, ge=0.0, lt=0.5, description="MAF 过滤下限")
    n_pc: int = Field(3, ge=1, le=10, description="群体结构主成分数")
    sig_level: float = Field(0.05, gt=0.0, le=1.0, description="显著性水平")
    qtl_window: int = Field(1_000_000, ge=1, description="QTL 合并窗口(bp)")
    env_dim: str = Field("site", description="环境维度：site(trial_study_id) / year(evaluate_year)")
    model_config = ConfigDict(extra="ignore")


class GwasResultOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    dataset_id: int | None = None
    trait_id: int | None = None
    trait_code: str | None = None
    method: str | None = None
    n_individuals: int | None = None
    n_markers: int | None = None
    m_after_maf: int | None = None
    maf_min: float | None = None
    n_pc: int | None = None
    sig_level: float | None = None
    threshold_bonf: float | None = None
    n_sig_bonf: int | None = None
    n_sig_fdr: int | None = None
    n_qtl: int | None = None
    data_version: str | None = None
    input_hash: str | None = None
    engine_version: str | None = None
    remark: str | None = None
    created_time: datetime | None = None


class GwasSnpOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    gwas_result_id: int
    marker_id: int | None = None
    marker_name: str | None = None
    chromosome: str | None = None
    position: int | None = None
    maf: float | None = None
    effect: float | None = None
    se: float | None = None
    t_value: float | None = None
    p_value: float | None = None
    neg_log10p: float | None = None
    q_value: float | None = None
    sig_bonf: bool | None = None
    sig_fdr: bool | None = None


class QtlIn(BaseModel):
    """已知 QTL 录入（source=known）入参。"""
    trait_id: int | None = Field(None, description="性状 id（bre_trait）")
    chromosome: str | None = Field(None, description="染色体")
    start_bp: int | None = Field(None, description="区间起点(bp)")
    end_bp: int | None = Field(None, description="区间终点(bp)")
    peak_marker_id: int | None = Field(None, description="峰标记 id（bre_marker）")
    peak_p: float | None = Field(None, description="峰标记 p 值")
    n_markers: int | None = Field(None, description="区间显著标记数")
    effect: float | None = Field(None, description="峰效应")
    remark: str | None = Field(None, description="备注")
    model_config = ConfigDict(extra="ignore")


class QtlOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    trait_id: int | None = None
    chromosome: str | None = None
    start_bp: int | None = None
    end_bp: int | None = None
    peak_marker_id: int | None = None
    peak_marker_name: str | None = None
    peak_p: float | None = None
    n_markers: int | None = None
    effect: float | None = None
    source: str | None = None
    gwas_result_id: int | None = None
    remark: str | None = None
    created_time: datetime | None = None


class MasPanelIn(BaseModel):
    """MAS 标记辅助选择面板入参。"""
    panel_name: str = Field(..., description="面板名称")
    trait_id: int | None = Field(None, description="目标性状 id（bre_trait）")
    remark: str | None = Field(None, description="备注")
    model_config = ConfigDict(extra="ignore")


class MasPanelOut(BaseModel):
    model_config = ConfigDict(from_attributes=True)
    id: int
    panel_name: str | None = None
    trait_id: int | None = None
    n_markers: int | None = None
    remark: str | None = None
    created_time: datetime | None = None


class MasPanelMarkerIn(BaseModel):
    """面板标记（有利剂量 + 方向 + 基因作用模式）。"""
    marker_id: int | None = Field(None, description="标记 id（bre_marker）")
    favorable_dose: int | None = Field(None, description="有利剂量(0/1/2)")
    effect: float | None = Field(None, description="效应(参考)")
    direction: str | None = Field("high", description="high=剂量≥favorable_dose命中 / low=剂量≤命中")
    mode: str | None = Field("additive", description="基因作用模式：additive/dominance/recessive/allele/haplotype")
    favorable_allele: str | None = Field(None, description="有利等位(allele/haplotype 模式用，SSR 等位索引)")
    haplotype_group: str | None = Field(None, description="单倍型组(haplotype 模式：同组标记全命中才计 1)")
    model_config = ConfigDict(extra="ignore")


class MasPanelSetMarkersIn(BaseModel):
    """面板标记全量替换入参。"""
    markers: list[MasPanelMarkerIn] = Field(default_factory=list, description="面板标记列表")
    model_config = ConfigDict(extra="ignore")