from analytics.curve_trades import on_the_run
from analytics.index_data import get_index_quotes, index_returns
from utils.db import serenitas_engine, dawn_engine

import pandas as pd
import math
import datetime


def hist_var(portf, index_type="IG", quantile=0.05, years=5):
    df = index_returns(
        index=index_type, years=years, tenor=["3yr", "5yr", "7yr", "10yr"]
    )
    df = df.reset_index(["index"], drop=True).reorder_levels(
        ["date", "series", "tenor"]
    )
    returns = df.spread_return.dropna().reset_index("series")
    returns["dist_on_the_run"] = returns.groupby("date")["series"].transform(
        lambda x: x.max() - x
    )
    del returns["series"]
    returns = returns.set_index("dist_on_the_run", append=True).unstack("tenor")
    returns.columns = returns.columns.droplevel(0)
    portf.reset_pv()

    otr = on_the_run(index_type)
    spreads = pd.DataFrame(
        {
            "spread": portf.spread,
            "tenor": [ind.tenor for ind in portf.indices],
            "dist_on_the_run": [otr - ind.series for ind in portf.indices],
        }
    )
    spreads = spreads.set_index(["dist_on_the_run", "tenor"])
    r = []
    for k, g in returns.groupby(level="date", as_index=False):
        shocks = g.reset_index("date", drop=True).stack("tenor")
        shocks.name = "shocks"
        portf.spread = spreads.spread * (1 + spreads.join(shocks).shocks)
        r.append((k, portf.pnl))
    pnl = pd.DataFrame.from_records(r, columns=["date", "pnl"], index=["date"])
    return pnl.quantile(quantile) * math.sqrt(20)