import pandas as pd
import numpy as np
import metrics as mt

# metrics I'm not inclined to include in the paper
#               'DecT':get_declin_tilde,\
#               'BDecT':get_bdec_tilde,\

# all of the metrics we're considering in this paper
# note that Lop and EVW include corresponding statistical data
# see populate_metrics function for how this is recorded
metric_dict = {'Dec': mt.get_declin,\
               'BDec':mt.get_bdec,\
               'EG':mt.get_EG,\
               'LG':mt.get_EG_loss_only,\
               'DG':mt.get_EG_difference,\
               'SG':mt.get_EG_surplus_only,\
               'VC1':mt.get_EG_vote_centric_one,\
               'VC2':mt.get_EG_vote_centric_two,\
               'Bias':mt.get_bias,\
               'MM':mt.get_mean_median,\
               'Lop':mt.get_lop_list,\
               'EVW':mt.get_evw}

# exclude evw
m_noevw = dict(metric_dict)
del m_noevw['EVW']

# exclude evw
m_noevw_nosg = dict(m_noevw)
del m_noevw_nosg['SG']

# Todo: Rewrite so it's clearer which metrics not being included in this
m_list = {'Dec': mt.get_declin,\
          'BDec':mt.get_bdec,\
          'EG':mt.get_EG,\
          'LG':mt.get_EG_loss_only,\
          'DG':mt.get_EG_difference,\
          'SG':mt.get_EG_surplus_only,\
          'VC1':mt.get_EG_vote_centric_one,\
          'VC2':mt.get_EG_vote_centric_two,\
          'Bias':mt.get_bias,\
          'MM':mt.get_mean_median,\
          'Lop':mt.get_lop_list}

rescale_suffix = '-R'
absolute_value_suffix = '-abs'

rms_cols = [x + rescale_suffix for x in ['EG','BDec','MM','Bias']]

#####################################################################
# general code for working with the metrics dataframes
#####################################################################

#############################################################################
def make_df(elecs):
    """ make a pandas dataframe from elections for holding metric values
        populate with basic data about the election
    """
    elecdict = {'elecs': [x for x in elecs.keys()]}
    df = pd.DataFrame(elecdict)

    # This syntax is okay since df.elecs returns a series - only one axis to work with.
    df['year'] = df.elecs.apply(lambda x: x[:4])
    df['state'] = df.elecs.apply(lambda x: x[5:7])
    df['house'] = df.elecs.apply(lambda x: x[-2:] == '11')
    df['N'] = df.elecs.apply(lambda x: elecs[x].Ndists)

    # There's a MN election in database with no data - ignore it.
    df = df[df['N'] > 0]

    df['Davg'] = df.elecs.apply(lambda x: np.mean(elecs[x].demfrac))
    df['SeatDiff'] = df.elecs.apply(lambda x: len(filter(lambda x: x > 0.5, elecs[x].demfrac)) - elecs[x].Ndists*1.0/2)
    df['Sfrac'] = df.elecs.apply(lambda x: len(filter(lambda x: x > 0.5, elecs[x].demfrac))*1.0/elecs[x].Ndists)
    # is the election competitive?
    df['comp'] = df.elecs.apply(lambda x: 0.45 <= np.mean(elecs[x].demfrac) <= 0.55)
    # did each party win at least one seat?
    df['nosw'] = df.elecs.apply(lambda x: 0 < len(filter(lambda y: y > 0.5, elecs[x].demfrac)) < elecs[x].Ndists)
    return df

def populate_metrics(df,mdict,elecs):
    """ populate dataframe with metric values
    """
    for fkey in mdict.keys():
        df[fkey] = df.elecs.apply(lambda x: mdict[fkey](elecs[x].demfrac))

    # make adjustments for columns corresponding to "tests" (rather than measures)
    df['Lop-t'] = df.Lop.apply(lambda x: x[1]) # value of t-statistic
    df['Lop-p'] = df.Lop.apply(lambda x: x[2]) # p-value
    df['Lop-sig'] = df.Lop.apply(lambda x: x[3] < 0.5) # whether statistically significant
    df['Lop'] = df.Lop.apply(lambda x: x[0]) # difference between winning vote shares

    # make adjustments for EVW
    df['EVW-sig'] = df.EVW.apply(lambda x: x[1])
    df['EVW'] = df.EVW.apply(lambda x: x[0])

    # now that we've fixed things up
    for fkey in mdict.keys():
        df[fkey + absolute_value_suffix] = df[fkey].apply(lambda x: abs(x))

    return df

def rescale_metrics(df,mdict):
    """ rescale all metrics so standard deviation is 1. Note, not shifting mean?
    """
    # This is here to get rid of a SettingWithCopyWarning
    df = df.copy()
    for x in mdict:
        df.loc[:,x + rescale_suffix] = df[x]/df[x].std()
    return df

def get_sigma(df,mdict):
    """ return a dictionary containing standard deviations for requested columns
    """
    nd = dict()
    for x in mdict.keys():
        nd[x] = df[x].std()
    return nd

def make_pairwise(df,measures):
    """ make pairwise comparisons for the specified measures
    """
    arr = []
    for i,m1 in enumerate(measures):
        for j,m2 in enumerate(measures):
            if j > i:
                df.loc[:,m1 + '-' + m2 + absolute_value_suffix] = abs(df.loc[:,m1 + rescale_suffix] - df.loc[:,m2 + rescale_suffix])
                arr.append(m1 + '-' + m2)
    return arr
                
def get_rms(row):
    return np.sqrt(sum([(row[x]-row['Avg'])**2 for x in rms_cols]))

def find_least_consensus(df,n=8):
    """ add a new column with rms error among given measures
    """
    df['Avg'] = df[rms_cols].apply(np.mean, axis='columns')
    df['RMS'] = df.apply(get_rms, axis='columns')

    # df.loc[:,'RMS'] = df.loc[:,measures])
    tdf = df[['elecs','RMS']].sort_values('RMS')
    return tdf.nlargest(n,'RMS').reset_index(drop=True)