import torch
import torch.nn as nn
import torch.nn.functional as F


class DynamicTanh(nn.Module):
    """
    Dynamic tanh activation with learnable scaling (alpha) and affine transformation (weight, bias).
    """

    def __init__(self, normalized_shape, channels_last=True, alpha_init_value=0.5):
        super().__init__()
        self.normalized_shape = normalized_shape
        self.alpha_init_value = alpha_init_value
        self.channels_last = channels_last

        # learnable scale for tanh
        self.alpha = nn.Parameter(torch.full((1,), alpha_init_value))
        # affine parameters
        self.weight = nn.Parameter(torch.ones(normalized_shape))
        self.bias = nn.Parameter(torch.zeros(normalized_shape))

    def forward(self, x):
        # scaled tanh
        x = torch.tanh(self.alpha * x)
        # affine transform
        if self.channels_last:
            x = x * self.weight + self.bias
        else:
            # channels_first: assume shape (B, C, H, W)
            x = x * self.weight[:, None, None] + self.bias[:, None, None]
        return x

    def extra_repr(self):
        return f"normalized_shape={self.normalized_shape}, alpha_init_value={self.alpha_init_value}, channels_last={self.channels_last}"


class DynamicGraphConstructor(nn.Module):
    def __init__(self, node_num, embed_dim):
        super().__init__()
        self.nodevec1 = nn.Parameter(
            torch.randn(node_num, embed_dim), requires_grad=True
        )
        self.nodevec2 = nn.Parameter(
            torch.randn(node_num, embed_dim), requires_grad=True
        )

    def forward(self):
        adj = torch.matmul(self.nodevec1, self.nodevec2.T)
        adj = F.relu(adj)
        adj = F.softmax(adj, dim=-1)
        return adj


class GraphConvBlock(nn.Module):
    def __init__(self, input_dim, output_dim):
        super().__init__()
        self.theta = nn.Linear(input_dim, output_dim)
        self.residual = input_dim == output_dim
        if not self.residual:
            self.res_proj = nn.Linear(input_dim, output_dim)

    def forward(self, x, adj):
        res = x
        x = torch.matmul(adj, x)
        x = self.theta(x)
        x = x + (res if self.residual else self.res_proj(res))
        return F.relu(x)


class MANBA_Block(nn.Module):
    """
    Multi-head attention + feed-forward network with DynamicTanh replacing LayerNorm.
    """

    def __init__(self, input_dim, hidden_dim):
        super().__init__()
        self.attn = nn.MultiheadAttention(
            embed_dim=input_dim, num_heads=4, batch_first=True
        )
        self.ffn = nn.Sequential(
            nn.Linear(input_dim, hidden_dim),
            nn.ReLU(),
            nn.Linear(hidden_dim, input_dim),
        )
        # replace LayerNorm with DynamicTanh
        self.norm1 = DynamicTanh(normalized_shape=input_dim, channels_last=True)
        self.norm2 = DynamicTanh(normalized_shape=input_dim, channels_last=True)

    def forward(self, x):
        # self-attention
        res = x
        x_attn, _ = self.attn(x, x, x)
        x = self.norm1(res + x_attn)
        # feed-forward
        res2 = x
        x_ffn = self.ffn(x)
        x = self.norm2(res2 + x_ffn)
        return x


class SandwichBlock(nn.Module):
    def __init__(self, num_nodes, embed_dim, hidden_dim):
        super().__init__()
        self.manba1 = MANBA_Block(hidden_dim, hidden_dim * 2)
        self.graph_constructor = DynamicGraphConstructor(num_nodes, embed_dim)
        self.gc = GraphConvBlock(hidden_dim, hidden_dim)
        self.manba2 = MANBA_Block(hidden_dim, hidden_dim * 2)

    def forward(self, h):
        h1 = self.manba1(h)
        adj = self.graph_constructor()
        h2 = self.gc(h1, adj)
        h3 = self.manba2(h2)
        return h3


class MLP(nn.Module):
    def __init__(self, in_dim, hidden_dims, out_dim, activation=nn.ReLU):
        super().__init__()
        dims = [in_dim] + hidden_dims + [out_dim]
        layers = []
        for i in range(len(dims) - 2):
            layers.append(nn.Linear(dims[i], dims[i + 1]))
            layers.append(activation())
        layers.append(nn.Linear(dims[-2], dims[-1]))
        self.net = nn.Sequential(*layers)

    def forward(self, x):
        return self.net(x)


class EXP(nn.Module):
    def __init__(self, args):
        super().__init__()
        self.horizon = args["horizon"]
        self.output_dim = args["output_dim"]
        self.seq_len = args.get("in_len", 12)
        self.hidden_dim = args.get("hidden_dim", 64)
        self.num_nodes = args["num_nodes"]
        self.embed_dim = args.get("embed_dim", 16)

        # discrete time embeddings
        self.time_slots = args.get("time_slots", 24 * 60 // args.get("time_slot", 5))
        self.time_embedding = nn.Embedding(self.time_slots, self.hidden_dim)
        self.day_embedding = nn.Embedding(7, self.hidden_dim)

        # input projection for flow history
        self.input_proj = MLP(
            in_dim=self.seq_len, hidden_dims=[self.hidden_dim], out_dim=self.hidden_dim
        )

        # two Sandwich blocks
        self.sandwich1 = SandwichBlock(self.num_nodes, self.embed_dim, self.hidden_dim)
        self.sandwich2 = SandwichBlock(self.num_nodes, self.embed_dim, self.hidden_dim)

        # output projection
        self.out_proj = MLP(
            in_dim=self.hidden_dim,
            hidden_dims=[2 * self.hidden_dim],
            out_dim=self.horizon * self.output_dim,
        )

    def forward(self, x):
        """
        x: (B, T, N, D_total) where
           x[...,0]=flow, x[...,1]=time_in_day (scaled), x[...,2]=day_in_week
        """
        x_flow = x[..., 0]  # (B, T, N)
        x_time = x[..., 1]  # (B, T, N)
        x_day = x[..., 2]  # (B, T, N)

        B, T, N = x_flow.shape
        assert T == self.seq_len, "Input sequence length mismatch"

        # project flow history
        x_flat = x_flow.permute(0, 2, 1).reshape(B * N, T)
        h0 = self.input_proj(x_flat).view(B, N, self.hidden_dim)

        # time embeddings at last step
        t_idx = (x_time[:, -1, :] * (self.time_slots - 1)).long()
        d_idx = x_day[:, -1, :].long()
        time_emb = self.time_embedding(t_idx)
        day_emb = self.day_embedding(d_idx)

        # inject time features
        h0 = h0 + time_emb + day_emb

        # Sandwich + residuals
        h1 = self.sandwich1(h0) + h0
        h2 = self.sandwich2(h1)

        # output
        out = self.out_proj(h2)
        out = out.view(B, N, self.horizon, self.output_dim)
        out = out.permute(0, 2, 1, 3)
        return out


# Example usage:
# args = {'horizon':12, 'output_dim':1, 'num_nodes':170}
# model = EXP(args)
# print(model)