import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.nn.utils.parametrizations import weight_norm, spectral_norm

# -------------------------------------------------------------------
# 1. Multi-Period Discriminator (MPD)
#    Captures periodic structures (pitch/timbre) by folding audio.
# -------------------------------------------------------------------

class DiscriminatorP(nn.Module):
    def __init__(self, period, kernel_size=5, stride=3, use_spectral_norm=False):
        super(DiscriminatorP, self).__init__()
        self.period = period
        self.use_spectral_norm = use_spectral_norm

        # Use spectral_norm for stability, or weight_norm for performance
        norm_f = spectral_norm if use_spectral_norm else weight_norm

        # We use 2D convs because we "fold" the 1D audio into 2D (Period x Time)
        self.convs = nn.ModuleList([
            norm_f(nn.Conv2d(1, 32, (kernel_size, 1), (stride, 1), padding=(2, 0))),
            norm_f(nn.Conv2d(32, 128, (kernel_size, 1), (stride, 1), padding=(2, 0))),
            norm_f(nn.Conv2d(128, 512, (kernel_size, 1), (stride, 1), padding=(2, 0))),
            norm_f(nn.Conv2d(512, 1024, (kernel_size, 1), (stride, 1), padding=(2, 0))),
            norm_f(nn.Conv2d(1024, 1024, (kernel_size, 1), 1, padding=(2, 0))),
        ])

        self.conv_post = norm_f(nn.Conv2d(1024, 1, (3, 1), 1, padding=(1, 0)))

    def forward(self, x):
        fmap = []

        # 1d to 2d conversion: [B, C, T] -> [B, C, T/P, P]
        b, c, t = x.shape
        if t % self.period != 0: # Pad if not divisible by period
            n_pad = self.period - (t % self.period)
            x = F.pad(x, (0, n_pad), "reflect")
            t = t + n_pad

        x = x.view(b, c, t // self.period, self.period)

        for l in self.convs:
            x = l(x)
            x = F.leaky_relu(x, 0.1)
            fmap.append(x) # Store feature map for Feature Matching Loss

        x = self.conv_post(x)
        fmap.append(x)

        # Flatten back to 1D for score
        x = torch.flatten(x, 1, -1)

        return x, fmap


class MultiPeriodDiscriminator(nn.Module):
    def __init__(self, periods=[2, 3, 5, 7, 11]):
        super(MultiPeriodDiscriminator, self).__init__()
        self.discriminators = nn.ModuleList([
            DiscriminatorP(p) for p in periods
        ])

    def forward(self, y, y_hat):
        y_d_rs = [] # Real scores
        y_d_gs = [] # Generated (Fake) scores
        fmap_rs = [] # Real feature maps
        fmap_gs = [] # Generated (Fake) feature maps

        for i, d in enumerate(self.discriminators):
            y_d_r, fmap_r = d(y)
            y_d_g, fmap_g = d(y_hat)

            y_d_rs.append(y_d_r)
            fmap_rs.append(fmap_r)
            y_d_gs.append(y_d_g)
            fmap_gs.append(fmap_g)

        return y_d_rs, y_d_gs, fmap_rs, fmap_gs


# -------------------------------------------------------------------
# 2. Multi-Scale Discriminator (MSD)
#    Captures structure at different audio resolutions (raw, x0.5, x0.25).
# -------------------------------------------------------------------

class DiscriminatorS(nn.Module):
    def __init__(self, use_spectral_norm=False):
        super(DiscriminatorS, self).__init__()
        norm_f = spectral_norm if use_spectral_norm else weight_norm

        # Standard 1D Convolutions with large receptive field
        self.convs = nn.ModuleList([
            norm_f(nn.Conv1d(1, 16, 15, 1, padding=7)),
            norm_f(nn.Conv1d(16, 64, 41, 4, groups=4, padding=20)),
            norm_f(nn.Conv1d(64, 256, 41, 4, groups=16, padding=20)),
            norm_f(nn.Conv1d(256, 1024, 41, 4, groups=64, padding=20)),
            norm_f(nn.Conv1d(1024, 1024, 41, 4, groups=256, padding=20)),
            norm_f(nn.Conv1d(1024, 1024, 5, 1, padding=2)),
        ])
        self.conv_post = norm_f(nn.Conv1d(1024, 1, 3, 1, padding=1))

    def forward(self, x):
        fmap = []
        for l in self.convs:
            x = l(x)
            x = F.leaky_relu(x, 0.1)
            fmap.append(x)
        x = self.conv_post(x)
        fmap.append(x)
        x = torch.flatten(x, 1, -1)
        return x, fmap


class MultiScaleDiscriminator(nn.Module):
    def __init__(self):
        super(MultiScaleDiscriminator, self).__init__()
        # 3 Scales: Original, Downsampled x2, Downsampled x4
        self.discriminators = nn.ModuleList([
            DiscriminatorS(use_spectral_norm=True),
            DiscriminatorS(),
            DiscriminatorS(),
        ])
        self.meanpools = nn.ModuleList([
            nn.AvgPool1d(4, 2, padding=2),
            nn.AvgPool1d(4, 2, padding=2)
        ])

    def forward(self, y, y_hat):
        y_d_rs = []
        y_d_gs = []
        fmap_rs = []
        fmap_gs = []

        for i, d in enumerate(self.discriminators):
            if i != 0:
                # Downsample input for subsequent discriminators
                y = self.meanpools[i-1](y)
                y_hat = self.meanpools[i-1](y_hat)

            y_d_r, fmap_r = d(y)
            y_d_g, fmap_g = d(y_hat)

            y_d_rs.append(y_d_r)
            fmap_rs.append(fmap_r)
            y_d_gs.append(y_d_g)
            fmap_gs.append(fmap_g)

        return y_d_rs, y_d_gs, fmap_rs, fmap_gs


# -------------------------------------------------------------------
# 3. Master Wrapper
#    Combines MPD and MSD into one class to fit your training script.
# -------------------------------------------------------------------

class SISUDiscriminator(nn.Module):
    def __init__(self):
        super(SISUDiscriminator, self).__init__()
        self.mpd = MultiPeriodDiscriminator()
        self.msd = MultiScaleDiscriminator()

    def forward(self, y, y_hat):
        # Return format:
        # scores_real, scores_fake, features_real, features_fake

        # Run Multi-Period
        mpd_y_d_rs, mpd_y_d_gs, mpd_fmap_rs, mpd_fmap_gs = self.mpd(y, y_hat)

        # Run Multi-Scale
        msd_y_d_rs, msd_y_d_gs, msd_fmap_rs, msd_fmap_gs = self.msd(y, y_hat)

        # Combine all results
        return (
            mpd_y_d_rs + msd_y_d_rs, # All real scores
            mpd_y_d_gs + msd_y_d_gs, # All fake scores
            mpd_fmap_rs + msd_fmap_rs, # All real feature maps
            mpd_fmap_gs + msd_fmap_gs  # All fake feature maps
        )