import pandas as pd
import networkx as nx
import numpy as np

# CARREGANDO DATAFRAME
df_rio = pd.read_csv('./dataset.csv')

df_rio['original_id'] = df_rio['original'] + '/' + df_rio['original_photo']
df_rio['compare_id'] = df_rio['compare'] + '/' + df_rio['compare_photo']

# CRIANDO GRAFO
G = nx.Graph()

for index, row in df_rio.iterrows():
    if row['original_id'] not in G:
        G.add_node(row['original_id'])
    if row['compare_id'] not in G:
        G.add_node(row['compare_id'])

    weight = row['distance']

    G.add_edge(row['original_id'], row['compare_id'], weight=weight)


def get_graph():
    return G


def get_page_rank_graph():
    # CRIANDO GRAFO PARA PAGE RANKING
    G_pr = nx.Graph()

    for index, row in df_rio.iterrows():
        if row['original_id'] not in G_pr:
            G_pr.add_node(row['original_id'])
        if row['compare_id'] not in G_pr:
            G_pr.add_node(row['compare_id'])

        weight = np.abs(row['distance'] - 1)

    G_pr.add_edge(row['original_id'], row['compare_id'], weight=weight)
    return G_pr

# ALGORITMOS DE RECOMENDAÇÃO


def recommend_dijkstra(graph, original_id):
    distances, paths = nx.single_source_dijkstra(
        graph, original_id, weight='weight')

    recommended_paths = sorted(
        paths.items(), key=lambda item: distances[item[0]])

    return recommended_paths


def recommend_floyd_warshall(graph, original_id):
    distances = nx.floyd_warshall_numpy(graph, weight='weight')

    nodes = list(graph.nodes())
    original_index = nodes.index(original_id)

    recommended_indices = sorted(
        range(len(nodes)), key=lambda i: distances[original_index][i])
    recommended_paths = [(nodes[i], distances[original_index][i])
                         for i in recommended_indices]

    return recommended_paths


def recommend_page_rank(graph, start_node):
    personalization = {node: 0 for node in graph.nodes()}
    personalization[start_node] = 1

    personalized_pagerank_scores = nx.pagerank(
        graph, personalization=personalization, weight='weight')
    recommended_paths = sorted(
        personalized_pagerank_scores.items(), key=lambda item: item[1], reverse=True)

    return recommended_paths