import os
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt


# ============================================================
# 設定
# ============================================================

INPUT_FILE = "results/trade_base.csv"
OUTPUT_DIR = "results"

os.makedirs(
    OUTPUT_DIR,
    exist_ok=True
)

D_LOWER = 0.60
D_UPPER = 1.00


# ============================================================
# 出力ファイル
# ============================================================

SUMMARY_CSV = os.path.join(
    OUTPUT_DIR,
    "d_market_regime_win_loss_summary.csv"
)

YEAR_CSV = os.path.join(
    OUTPUT_DIR,
    "d_market_regime_year_analysis.csv"
)

ATR_BOX_GRAPH = os.path.join(
    OUTPUT_DIR,
    "d_atr_win_loss_boxplot.png"
)

BB_BOX_GRAPH = os.path.join(
    OUTPUT_DIR,
    "d_bb_width_win_loss_boxplot.png"
)

ATR_YEAR_GRAPH = os.path.join(
    OUTPUT_DIR,
    "d_yearly_atr_median.png"
)

BB_YEAR_GRAPH = os.path.join(
    OUTPUT_DIR,
    "d_yearly_bb_width_median.png"
)


# ============================================================
# 読み込み
# ============================================================

df = pd.read_csv(
    INPUT_FILE
)

print("=" * 60)
print("D MARKET REGIME ANALYSIS")
print("=" * 60)

print(
    f"\n読み込み件数: {len(df)}"
)


# ============================================================
# 列確認
# ============================================================

required_columns = [
    "entry_time",
    "entry_band_distance_pips",
    "entry_atr14_pips",
    "profit",
]

missing_columns = [
    column
    for column in required_columns
    if column not in df.columns
]

if missing_columns:

    raise ValueError(
        "必要な列がありません: "
        + ", ".join(missing_columns)
    )


# ============================================================
# BB幅候補を探す
# ============================================================

bb_candidates = [
    column
    for column in df.columns
    if (
        "bb" in column.lower()
        and
        (
            "width" in column.lower()
            or
            "bandwidth" in column.lower()
        )
    )
]

print("\nBB幅候補列:")
print(bb_candidates)


if len(bb_candidates) == 0:

    BB_WIDTH_COLUMN = None

    print(
        "\nBB幅を表す列が見つかりませんでした。"
    )

    print(
        "今回はATR分析のみ実行します。"
    )

else:

    BB_WIDTH_COLUMN = (
        bb_candidates[0]
    )

    print(
        "\n使用するBB幅列:",
        BB_WIDTH_COLUMN
    )


# ============================================================
# 型変換
# ============================================================

df["entry_time"] = pd.to_datetime(
    df["entry_time"],
    errors="coerce"
)


numeric_columns = [
    "entry_band_distance_pips",
    "entry_atr14_pips",
    "profit",
]


if BB_WIDTH_COLUMN is not None:

    numeric_columns.append(
        BB_WIDTH_COLUMN
    )


for column in numeric_columns:

    df[column] = pd.to_numeric(
        df[column],
        errors="coerce"
    )


# ============================================================
# 基本データ
# ============================================================

analysis_df = df[
    df["entry_time"].notna()
    &
    df["entry_band_distance_pips"].notna()
    &
    df["entry_atr14_pips"].notna()
    &
    df["profit"].notna()
    &
    (
        df["entry_atr14_pips"] > 0
    )
].copy()


analysis_df[
    "relative_band_distance"
] = (
    analysis_df[
        "entry_band_distance_pips"
    ]
    /
    analysis_df[
        "entry_atr14_pips"
    ]
)


analysis_df[
    "relative_band_distance"
] = (
    analysis_df[
        "relative_band_distance"
    ]
    .replace(
        [
            np.inf,
            -np.inf
        ],
        np.nan
    )
)


analysis_df = analysis_df[
    analysis_df[
        "relative_band_distance"
    ].notna()
].copy()


analysis_df[
    "year"
] = (
    analysis_df[
        "entry_time"
    ].dt.year
)


# ============================================================
# D抽出
# ============================================================

d_df = analysis_df[
    analysis_df[
        "relative_band_distance"
    ].between(
        D_LOWER,
        D_UPPER,
        inclusive="both"
    )
].copy()


d_df[
    "result"
] = np.where(
    d_df["profit"] > 0,
    "WIN",
    "LOSS"
)


# 0円取引が存在する場合は除外
d_df = d_df[
    d_df["profit"] != 0
].copy()


print(
    f"\nD取引数: {len(d_df)}"
)

print(
    "WIN:",
    (
        d_df["result"]
        ==
        "WIN"
    ).sum()
)

print(
    "LOSS:",
    (
        d_df["result"]
        ==
        "LOSS"
    ).sum()
)


# ============================================================
# 要約統計
# ============================================================

def distribution_stats(
    source_df,
    column
):

    series = (
        source_df[column]
        .dropna()
    )

    if len(series) == 0:

        return {
            "count": 0,
            "mean": np.nan,
            "median": np.nan,
            "std": np.nan,
            "q25": np.nan,
            "q75": np.nan,
            "min": np.nan,
            "max": np.nan,
        }

    return {
        "count":
            len(series),

        "mean":
            series.mean(),

        "median":
            series.median(),

        "std":
            series.std(),

        "q25":
            series.quantile(0.25),

        "q75":
            series.quantile(0.75),

        "min":
            series.min(),

        "max":
            series.max(),
    }


summary_rows = []


analysis_variables = {
    "ATR14":
        "entry_atr14_pips"
}


if BB_WIDTH_COLUMN is not None:

    analysis_variables[
        "BB_WIDTH"
    ] = BB_WIDTH_COLUMN


for variable_name, column in (
    analysis_variables.items()
):

    for result in [
        "WIN",
        "LOSS"
    ]:

        subset = d_df[
            d_df["result"]
            ==
            result
        ]

        stats = distribution_stats(
            subset,
            column
        )

        stats[
            "variable"
        ] = variable_name

        stats[
            "result"
        ] = result

        summary_rows.append(
            stats
        )


summary_df = pd.DataFrame(
    summary_rows
)


summary_columns = [
    "variable",
    "result",
    "count",
    "mean",
    "median",
    "std",
    "q25",
    "q75",
    "min",
    "max",
]


summary_df = summary_df[
    summary_columns
]


summary_df.to_csv(
    SUMMARY_CSV,
    index=False
)


# ============================================================
# 年別分析
# ============================================================

year_rows = []


for year, year_df in d_df.groupby(
    "year"
):

    row = {
        "year":
            year,

        "trades":
            len(year_df),

        "total_profit":
            year_df[
                "profit"
            ].sum(),

        "avg_profit":
            year_df[
                "profit"
            ].mean(),

        "atr_mean":
            year_df[
                "entry_atr14_pips"
            ].mean(),

        "atr_median":
            year_df[
                "entry_atr14_pips"
            ].median(),
    }


    if BB_WIDTH_COLUMN is not None:

        row[
            "bb_width_mean"
        ] = (
            year_df[
                BB_WIDTH_COLUMN
            ].mean()
        )

        row[
            "bb_width_median"
        ] = (
            year_df[
                BB_WIDTH_COLUMN
            ].median()
        )


    year_rows.append(
        row
    )


year_df = pd.DataFrame(
    year_rows
).sort_values(
    "year"
)


year_df.to_csv(
    YEAR_CSV,
    index=False
)


# ============================================================
# ATR：WIN / LOSS 箱ひげ図
# ============================================================

win_atr = (
    d_df.loc[
        d_df["result"] == "WIN",
        "entry_atr14_pips"
    ]
    .dropna()
)


loss_atr = (
    d_df.loc[
        d_df["result"] == "LOSS",
        "entry_atr14_pips"
    ]
    .dropna()
)


plt.figure(
    figsize=(9, 6)
)


plt.boxplot(
    [
        win_atr,
        loss_atr
    ],
    labels=[
        "WIN",
        "LOSS"
    ],
    showfliers=True
)


plt.title(
    "D: ATR Distribution - WIN vs LOSS"
)

plt.xlabel(
    "Trade Result"
)

plt.ylabel(
    "ATR14 (pips)"
)

plt.grid(
    axis="y",
    alpha=0.3
)

plt.tight_layout()

plt.savefig(
    ATR_BOX_GRAPH,
    dpi=150
)

plt.close()


# ============================================================
# BB幅：WIN / LOSS 箱ひげ図
# ============================================================

if BB_WIDTH_COLUMN is not None:

    win_bb = (
        d_df.loc[
            d_df["result"] == "WIN",
            BB_WIDTH_COLUMN
        ]
        .dropna()
    )


    loss_bb = (
        d_df.loc[
            d_df["result"] == "LOSS",
            BB_WIDTH_COLUMN
        ]
        .dropna()
    )


    plt.figure(
        figsize=(9, 6)
    )


    plt.boxplot(
        [
            win_bb,
            loss_bb
        ],
        labels=[
            "WIN",
            "LOSS"
        ],
        showfliers=True
    )


    plt.title(
        "D: BB Width Distribution - WIN vs LOSS"
    )

    plt.xlabel(
        "Trade Result"
    )

    plt.ylabel(
        BB_WIDTH_COLUMN
    )

    plt.grid(
        axis="y",
        alpha=0.3
    )

    plt.tight_layout()

    plt.savefig(
        BB_BOX_GRAPH,
        dpi=150
    )

    plt.close()


# ============================================================
# 年別ATR中央値
# ============================================================

plt.figure(
    figsize=(12, 7)
)


plt.plot(
    year_df["year"],
    year_df["atr_median"],
    marker="o"
)


plt.title(
    "D: Yearly Median ATR"
)

plt.xlabel(
    "Year"
)

plt.ylabel(
    "Median ATR14 (pips)"
)

plt.grid(
    alpha=0.3
)

plt.tight_layout()

plt.savefig(
    ATR_YEAR_GRAPH,
    dpi=150
)

plt.close()


# ============================================================
# 年別BB幅中央値
# ============================================================

if BB_WIDTH_COLUMN is not None:

    plt.figure(
        figsize=(12, 7)
    )


    plt.plot(
        year_df["year"],
        year_df[
            "bb_width_median"
        ],
        marker="o"
    )


    plt.title(
        "D: Yearly Median BB Width"
    )

    plt.xlabel(
        "Year"
    )

    plt.ylabel(
        BB_WIDTH_COLUMN
    )

    plt.grid(
        alpha=0.3
    )

    plt.tight_layout()

    plt.savefig(
        BB_YEAR_GRAPH,
        dpi=150
    )

    plt.close()


# ============================================================
# 表示
# ============================================================

print("\n")
print("=" * 60)
print("WIN / LOSS DISTRIBUTION")
print("=" * 60)

print(
    summary_df.to_string(
        index=False
    )
)


print("\n")
print("=" * 60)
print("YEAR ANALYSIS")
print("=" * 60)

print(
    year_df.to_string(
        index=False
    )
)


print("\n")
print("=" * 60)
print("OUTPUT FILES")
print("=" * 60)

print(
    SUMMARY_CSV
)

print(
    YEAR_CSV
)

print(
    ATR_BOX_GRAPH
)

print(
    ATR_YEAR_GRAPH
)


if BB_WIDTH_COLUMN is not None:

    print(
        BB_BOX_GRAPH
    )

    print(
        BB_YEAR_GRAPH
    )