import multiprocessing
import os
import sys
import time
import traceback

sys.path.append(os.path.realpath("."))

from gensim.models import Word2Vec, word2vec
from gm_rpcd.all import bind
from utils.es import es_scan
from utils.files import DATA_PATH, MODEL_PATH
from utils.message import (send_msg_to_dingtalk, send_performance_msg_to_dingtalk)

from word_vector.tractate import TRACTATE_CLICK_IDS, TRACTATE_CLICK_IDS_MODEL

model_output_name = "w2v_model"
model_path = os.path.join(MODEL_PATH, model_output_name)
try:
    WORD2VEC_MODEL = word2vec.Word2Vec.load(model_path)
except Exception as e:
    print(e)


class W2vSentences:
    def __init__(self, f_name):
        self.f_name = f_name

    def __iter__(self):
        with open(self.f_name, mode="r", encoding="utf-8", errors="ignore") as f:
            for line in f.readlines():
                yield line.split()


def w2v_train(f_name, model_output_name):
    input_file = os.path.join(DATA_PATH, f_name)
    print("input: " + input_file)
    sentences = W2vSentences(input_file)
    w2v_model = word2vec.Word2Vec(sentences, min_count=2, workers=2, size=100, window=10)
    model_path = os.path.join(MODEL_PATH, model_output_name)
    print("output: " + model_path)
    w2v_model.save(model_path)


@bind("strategy_embedding/word_vector/word_similarity")
def word_similarity(word):
    try:
        return WORD2VEC_MODEL.wv.most_similar(word)
    except Exception as e:
        send_msg_to_dingtalk(str(traceback.format_exc()))
        return []


def get_user_portrait_projects(score_limit=5):
    """
    return: {
      '6231F098-9E72-448E-B8D2-19FCB9687005': ['鼻综合', '玻尿酸填充面部'],
      '862538030266882': ['吸脂瘦脸', '吸脂瘦全身']
    }
    """
    es_res = es_scan("device", {}, rw=None)
    count = 0
    res = {}
    for i in es_res:
        count += 1
        print(count)
        source = i["_source"]
        device_id = source.get("device_id", "")
        projects = [i["name"] for i in source.get("projects", []) if i["score"] >= score_limit]
        if projects:
            res[device_id] = projects
    return res


def projects_item2vec(score_limit=5):
    user_dict = get_user_portrait_projects(score_limit=score_limit)
    # TODO if not redis.get user_dict:
    projects = list(user_dict.values())
    model = Word2Vec(projects, hs=0, min_count=3, workers=multiprocessing.cpu_count(), iter=10)
    print(model)
    print(len(projects))
    for word in ["鼻综合", "吸脂瘦脸"]:
        print(model.wv.most_similar(word, topn=5))
    return model


# item2vec
@bind("strategy_embedding/word_vector/tractate_item2vec")
def clicked_tractate_ids_item2vec_model(id, n=5):
    try:
        time_begin = time.time()
        res = TRACTATE_CLICK_IDS_MODEL.wv.most_similar(id, topn=n)
        time_end = time.time() - time_begin
        if time_end > 0.04:
            send_performance_msg_to_dingtalk("clicked_tractate_ids_item2vec_model id={} n={} cost {:.3f}ms".format(
                id, n, time_end * 1000))
        return res
    except KeyError as e:
        # send_msg_to_dingtalk("tractate_item2vec: " + str(e))
        return []
    except Exception as e:
        send_msg_to_dingtalk(str(traceback.format_exc()))
        return []


@bind("strategy_embedding/word_vector/tractate_item2vec_by_ids")
def clicked_tractate_ids_item2vec_model_by_ids(ids, n=5):
    """
    ids: ["417522", "179986", "75804", "84682", "400022"]
    return: [('77419', 0.8857042193412781),
             ('75761', 0.7772561311721802),
             ('84630', 0.9709808826446533),
             ('84671', 0.9621062278747559)]
    """
    try:
        time_begin = time.time()
        res = []
        for id in ids:
            if id in TRACTATE_CLICK_IDS:
                res.extend(TRACTATE_CLICK_IDS_MODEL.wv.most_similar(id, topn=n))
        time_end = time.time() - time_begin
        if time_end > 0.04:
            send_performance_msg_to_dingtalk("clicked_tractate_ids_item2vec_model_by_ids ids={} n={} cost {:.3f}ms".format(
                ids, n, time_end * 1000))
        return res
    except Exception as e:
        send_msg_to_dingtalk(str(traceback.format_exc()))
        return []


# if __name__ == "__main__":

# w2v_train("dispose_problem.txt", model_output_name)

# for i in ["双眼皮", "隆鼻"]:
#     print(word_similarity(i))

# save_clicked_tractate_ids_item2vec()

# for id in ["84375", "148764", "368399"]:
#     print(clicked_tractate_ids_item2vec_model(id, n=5))