import streamlit as st
import pandas as pd
import numpy as np
import torch
from transformers import AlbertTokenizer, AlbertModel
from sklearn.metrics.pairwise import cosine_similarity

# base is smaller, vs large
model_size='base'
tokenizer = AlbertTokenizer.from_pretrained('albert-' + model_size + '-v2')
model = AlbertModel.from_pretrained('albert-' + model_size + '-v2')

def get_embedding(input_text):
    encoded_input = tokenizer(input_text, return_tensors='pt')
    input_ids = encoded_input.input_ids
    #input_num_tokens = input_ids.shape[1]

    #print( "Number of input tokens: " + str(input_num_tokens))
    #print("Length of input: " + str(len(input_text)))

    list_of_tokens = tokenizer.convert_ids_to_tokens(input_ids.view(-1).tolist())

    print( "Tokens : " + ' '.join(list_of_tokens))
    with torch.no_grad():

        outputs = model(**encoded_input)
        last_hidden_states = outputs[0]
        sentence_embedding = torch.mean(last_hidden_states[0], dim=0)
        #sentence_embedding = output.last_hidden_state[0][0]
        return sentence_embedding.tolist()

st. set_page_config(layout="wide")
st.title('Upload the Address Dataset')

st.markdown('Upload an Excel file to view the data in a table.')

uploaded_file = st.file_uploader('Choose a file', type='xlsx')



if uploaded_file is not None:
    data_caqh = pd.read_excel(uploaded_file, sheet_name='CAQH', dtype=str)
    data_ndb = pd.read_excel(uploaded_file, sheet_name='NDB', dtype=str)

    # Data cleaning CAQH
    data_caqh['postalcode'] = data_caqh['postalcode'].astype(str).apply(lambda x: x[:5] + '-' + x[5:] if len(x) > 5 and not '-' in x else x)
    data_caqh['full-addr'] = data_caqh['address1'].astype(str) + ', ' \
                             + np.where(data_caqh['address2'].isnull(),  '' , data_caqh['address2'].astype(str))  \
                             + data_caqh['city'].astype(str) + ', '\
                             + data_caqh['state'].astype(str) + ', ' \
                             + data_caqh['postalcode'].astype(str)

    # Data cleaning NDB
    data_ndb['zip_pls_4_cd'] = data_ndb['zip_pls_4_cd'].astype(str).apply(lambda x: x if (x[-1] != '0' and x[-1] != '1') else '')

    data_ndb['zip_cd_zip_pls_4_cd'] = data_ndb['zip_cd'].astype(str) +\
                                      np.where( data_ndb['zip_pls_4_cd'] == '', '', '-' \
                                      + data_ndb['zip_pls_4_cd'].astype(str))

    data_ndb['full-addr'] = data_ndb['adr_ln_1_txt'].astype(str).str.strip() + ', ' \
                            + data_ndb['cty_nm'].astype(str) + ', ' \
                            + data_ndb['st_cd'].astype(str) + ', ' + data_ndb['zip_cd_zip_pls_4_cd']

    # App
    data_caqh['embedding'] = data_caqh['full-addr'].apply(get_embedding)
    data_ndb['embedding'] = data_ndb['full-addr'].apply(get_embedding)

    data_caqh['matched-addr'] = ''

    for i, row in data_caqh.iterrows():
        max_similarity = 0
        matched_row = None
        for j, ndb_row in data_ndb.iterrows():
            sim = cosine_similarity([row['embedding']], [ndb_row['embedding']])
            if sim > max_similarity:
                max_similarity = sim
                matched_row = ndb_row
        if max_similarity >= 0.98:
            data_caqh.at[i, 'matched-addr'] = matched_row['full-addr']
            data_caqh.at[i, 'similarity-score'] = max_similarity
        else:
            data_caqh.at[i, 'matched-addr'] = 'No Matches'

    # Drop columns not needed for display
    data_caqh.drop(columns=['embedding'], inplace=True)
    data_ndb.drop(columns=['embedding'], inplace=True)

    st.header('CAQH addresses and matches')
    st.dataframe(data_caqh, use_container_width=True)
    st.header('NDB data')
    st.dataframe(data_ndb, use_container_width=True)

    # calculate the embedding of each item.

    #st.dataframe(data_caqh)
    # Do some matching
    #data_caqh.loc[data_caqh['full-addr'] == '1000 Vale Terrace, Vista, CA, 92084', 'matched-addr'] = '456 Main St'
    #time.sleep(10)
    #st.dataframe(data_caqh)