'''
Created on Dec 16, 2014
@author: Brett Paufler
Copyright Brett Paufler
'''


import pandas as pd
import numpy as np
import trades
from trades import report
from trades import tdays
import matplotlib.pyplot as plt


def linear_graph(d1,d2):
    
    dF = trades.get_trades("corn")


    dF = dF[dF["Issue"] == "CZ2014"]
    dF = dF[dF["Date"] > d1]
    dF = dF[dF["Date"] < d2]
    
    #ADDS WEEKDAY COLUMNS (MON-FRI)
    for w in range(5):
        dF[tdays[w]] = dF["Settle"].where(dF["Date"].apply(lambda x: x.weekday()==w))


    gF = dF[["Date"] + list(tdays.values())]
    
    report(gF)
    
    plt.figure(figsize=(10,5))
    ax = gF.interpolate().plot(x="Date")
    fig = ax.get_figure()
    #fig.show()
    sN = "linear_graph_CZ-%s-%s.png" % (d1,d2)
    fig.savefig(sN)


def linear_graph_soybeans(d1=0,d2=0):
    
    
    dF = trades.get_trades("soybeans")
    if d1:
        dF = dF[dF["Issue"] == "SX2014"]
        dF = dF[dF["Date"] > d1]
        dF = dF[dF["Date"] < d2]
    
    #ADDS WEEKDAY COLUMNS (MON-FRI)
    for w in range(5):
        dF[tdays[w]] = dF["Settle"].where(dF["Date"].apply(lambda x: x.weekday()==w))


    gF = dF[["Date"] + list(tdays.values())]
    
    report(gF)
    
    plt.figure(figsize=(10,5))
    ax = gF.interpolate().plot(x="Date")
    fig = ax.get_figure()
    #fig.show()
    sN = "linear_graph_SX-%s-%s.png" % (d1,d2)
    fig.savefig(sN)

#linear_graph("2014-10-1","2014-10-31")
#linear_graph_soybeans("2014-10-1","2014-10-31")
'''
for d1,d2 in [("2014-10-1","2014-12-31"),
              ("2014-11-1","2014-11-30"),
              ("2014-1-1","2014-12-31"),
              ]:
    linear_graph_soybeans(d1,d2)
'''



#COMING ALONG NICELY    
def simple_weekday_average_bar_graph(commodity, years):
    '''
    sell price averaged over length of issue, per day
    '''
    dF = trades.by_issue_year(commodity, years)

    #ADDS WEEKLY COLUMNS
    for w in range(5):
        dF[tdays[w]] = dF["Settle"].where(dF["Date"].apply(lambda x: x.weekday()==w))
        
    
    #num non NaN, so can get an average
    numObs = []
    for w in range(5):
        numObs.append(dF[trades.tdays[w]].count() )
        
        
    dF = dF.fillna(0)
    
    wkDay = {}
    for w in range(5):
        wkDay[trades.tdays[w]] = sum(dF[trades.tdays[w]])/numObs[w]

    minV = min(wkDay.values())
    ave = dF["Settle"].mean()
    
    colors = []
    for w in range(5):
        wkDay[trades.tdays[w]] = wkDay[trades.tdays[w]] - ave
        if wkDay[trades.tdays[w]] > 0:
            colors.append("g")
        else:
            colors.append("r")
    print wkDay
    print minV
    
    print "wkDay %s" % str(wkDay)
    print "obsCount %s" % str(numObs)
    
    gF = pd.DataFrame(wkDay, index=[""])
    gF = gF[trades.weekdayList]
    
    trades.report(gF)

    if isinstance(years,int):
        graphTitle = "%s %s" % (commodity, str(years))
    else:
        graphTitle = "%s %s-%s"% (commodity, str(years[0]), str(years[-1]))
    (commodity,str(years))
    #Graphing Logic for 5 series data
    plt.figure()
    ax = gF.ix[0].plot(kind='bar',
            figsize=(10,5),
            legend=False,
            rot=0,#set_xticks=trades.weekdayList,
            color = colors,
            title = "Daily Sum Method\n%s" % graphTitle
            )
    plt.ylabel("Difference in Average Trade")
    
    sN = "simple_weekday_average_%s.png" % graphTitle.replace(" ","_")
    plt.savefig(sN)

'''         
for a in trades.graphList:
    simple_weekday_average_bar_graph(a[0],a[1])   
'''




def top_five_percent_bar_graph(commodity, years=0, top=0.05, best=True):
    '''
    sell price averaged over length of issue, per day
    if years not specified, uses all trades
    if year is full range, only uses issue years
    
    '''
    if years == 0:
        dF = trades.get_trades("corn")
    else:
        dF = trades.by_issue_year(commodity, years)
    
    dF = dF.sort("Settle")
    
    
    if isinstance(years,int):
        graphTitle = "%s %s" % (commodity, str(years))
    else:
        graphTitle = "%s %s-%s"% (commodity, str(years[0]), str(years[-1]))
    
    if best:
        dF = dF.tail(round(dF.shape[0] * top))
        sN = "best_trades_%d_%s.png" % (100*top, graphTitle.replace(" ","_"))
        titleText = "Best %d%% of Trades\n%s" % (100*top,graphTitle)
    else:
        dF = dF.head(round(dF.shape[0] * top))
        sN = "worst_trades_%d_%s.png" % (100*top, graphTitle.replace(" ","_"))
        titleText = "Worst %d%% of Trades\n%s" % (100*top,graphTitle)
    #FROM HERE< CUT AND PASTE OF ABOVE FUNCTION
    
    #ADDS WEEKLY COLUMNS
    for w in range(5):
        dF[tdays[w]] = dF["Settle"].where(dF["Date"].apply(lambda x: x.weekday()==w))
        
    
    #num non NaN, so can get an average
    numObs = []
    for w in range(5):
        numObs.append(dF[trades.tdays[w]].count() )
        
        
    dF = dF.fillna(0)
    
    wkDay = {}
    for w in range(5):
        wkDay[trades.tdays[w]] = sum(dF[trades.tdays[w]])/numObs[w]
        

    minV = min(wkDay.values())
    ave = dF["Settle"].mean()
    
    colors = []
    for w in range(5):
        wkDay[trades.tdays[w]] = wkDay[trades.tdays[w]] - ave
        if wkDay[trades.tdays[w]] > 0:
            colors.append("g")
        else:
            colors.append("r")
    print wkDay
    print minV
    
    print "wkDay %s" % str(wkDay)
    print "obsCount %s" % str(numObs)
    
    
    
    gF = pd.DataFrame(wkDay, index=[""])
    gF = gF[trades.weekdayList]
    
    trades.report(gF)


        
        
        
        
    #(commodity,str(years))
    #Graphing Logic for 5 series data
    plt.figure()
    ax = gF.ix[0].plot(kind='bar',
            figsize=(10,5),
            legend=False,
            rot=0,#set_xticks=trades.weekdayList,
            color = colors,
            title = titleText
            )
    plt.ylabel("Overall Difference")
    
    
    plt.savefig(sN)

'''  
for t in [0.01,0.05,0.10,0.25]:
    for c in trades.graphList:
        for b in [True,False]:
            top_five_percent_bar_graph(c[0], c[1], t, b)
'''
#top_five_percent_bar_graph("corn",range(1960,2014))
#top_five_percent_bar_graph("corn",2014)
#top_five_percent_bar_graph("corn")



def best_of_month(commodity, best=True):

    if commodity == "corn":
        dF = trades.by_issue_year("corn",range(1960,2015))
        start = 1960
        m = 13
    elif commodity == "soybeans":
        dF = trades.by_issue_year("soybeans",range(1970,2015))
        start = 1970
        m = 12
    else:
        raise Exception("Not a valid commodity type")

    #print dF.shape

    #print dF.shape
    trades.report(dF)


    day = [0,0,0,0,0]
    for year in range(start,2015):
        for month in range(1,m):
            
            
            tF = dF[dF.index.year == year]
            tF = tF[tF.index.month == month]

            tF = tF.sort("Settle")
            if best:
                #print "HERE"
                #print tF
                day[tF.tail(1).index.weekday] += 1
            else:
                day[tF.head(1).index.weekday] += 1
    
    ave = sum(day)/5.0
    aveDay = []
    colors = []
    for d in day:
        a = d - ave
        aveDay.append(a)
        if a > 0:
            colors.append("g")
        else:
            colors.append("r")
    
    gD = {"Monday":aveDay[0],
         "Tuesday":aveDay[1],
         "Wednesday":aveDay[2],
         "Thursday":aveDay[3],
         "Friday":aveDay[4]
         }
    
    print gD

    
    gF = pd.DataFrame(gD, index=[''])
    gF = gF[trades.weekdayList]
    
    print gF
    if best:
        titleText = "%s Best Days" %(commodity.capitalize())
        sN = "%s_Best_Days.png"%(commodity.capitalize())
    else:
        titleText = "%s Worst Days" %(commodity.capitalize())
        sN = "%s_Worst_Days.png"%(commodity.capitalize())
    #(commodity,str(years))
    #Graphing Logic for 5 series data
    plt.figure()
    ax = gF.ix[0].plot(kind='bar',
            figsize=(10,5),
            legend=False,
            rot=0,#set_xticks=trades.weekdayList,
            color = colors,
            title = titleText
            )
    plt.ylabel("Relative Rankings")
    #plt.ylabel("Overall Difference")
    
    
    plt.savefig(sN)


best_of_month("corn", True)
best_of_month("corn", False)
best_of_month("soybeans", True)
best_of_month("soybeans", False)






print "This Run Finished"