from fastapi import FastAPI
import uvicorn

import pandas as pd
from joblib import Parallel, delayed
from redfin import Redfin
import requests
requests.urllib3.disable_warnings()

app = FastAPI()


#Endpoints 
#Root endpoints
@app.get("/")
def root():
    return {"API": "Redfin Address Scrap"}


def red_fin_api(add):
    client = Redfin()
    response = client.search(add)
    
    try:
        url = response['payload']['exactMatch']['url']
        initial_info = client.initial_info(url)
    except:
        initial_info = add
    
    try:
         property_id = initial_info['payload']['propertyId']
         mls_data = client.below_the_fold(property_id)
    except:
        mls_data = add
    

    try:
        
        lat,lon=initial_info['payload']['latLong'].values()
        img=initial_info['payload']['preloadImageUrls'][0]
        
        # int_group=r[1]['payload']['amenitiesInfo']['superGroups'][0]['amenityGroups']
       
        ext_prop=mls_data['payload']['amenitiesInfo']['superGroups'][1]['amenityGroups'][0]['amenityEntries']
        ext_prop=pd.DataFrame(ext_prop)
        ext_prop['amenityValues']=[i[0] for i in ext_prop['amenityValues'].values]
        ext_prop2=ext_prop[['referenceName','amenityValues']].T
        ext_prop2.columns=ext_prop2.values[0]
        ext_prop3=ext_prop2.tail(1).reset_index(drop=1)

        df=pd.DataFrame(mls_data['payload']['publicRecordsInfo']['basicInfo'],index=[0]).drop(columns=['apn','propertyLastUpdatedDate','displayTimeZone'])
    
        df['Lat']=lat
        df['Lon']=lon
        # df['Image']=img
        df2=df.join(ext_prop3)
        df2.insert(0,'url',f'https://www.redfin.com{url}')
        
    except:
        df2=pd.DataFrame({'Missing':[1]})  
    
    df2.insert(0,'Address Input',add)  
    

    return df2


    
def catch_errors(addresses):
    try: 
        return red_fin_api(addresses)
    except:
        return pd.DataFrame({'Address Input':[addresses]})



    
    
    
@app.get('/Redfin_Address_Scrap')
async def predict(address_input: str):
    
    address_input_split = address_input.split(';')
    # results = process_multiple_address(address_input_split)
    results=Parallel(n_jobs=64, prefer="threads")(delayed(catch_errors)(i) for i in address_input_split)
    results = pd.concat(results).reset_index(drop=1)
    
    # cols_order=['Address Input',  'sqFtFinished', 'totalSqFt', 'yearBuilt', 'propertyTypeName', 'beds', 'baths', 'numStories',
    #             'url',
    #             'Lat', 'Lon']
    # cols_other=[i for i in results.columns if i not in cols_order ]
    # results=results[cols_order+cols_other].reset_index()
    
    results['index']=results['index']+1
    results.index=results.index+1
    results=results.rename(columns={'index':'Input Position'})
    prediction['yearBuilt']=results['yearBuilt'].fillna(0).astype(int).astype(str).replace('0','')

    return results.to_json()