How to extract past data on a custom query from Tweeter search APIs



I would like to collect some past data from Twitter on some search terms for a project. I applied for a developers account to get access to the data in past(>30 days). This is the first time I am using twitter API to collect data, I need some help and suggestions if I am doing something wrong. I tried three ways to collect data and was unable to fetch for >7 days actually data from 2012-2017 worth.

First, I tried using “searchtweets” which is Python wrapper for Twitter Premium and Enterprise Search APIs.
from searchtweets import ResultStream, gen_rule_payload, load_credentials

# first load your credentials

from searchtweets import collect_results
searchQuery1 = '"#JustinBieber" "#drunksinging" OR "#carelessDriving" OR "#outofcontrol" OR "#spitonfans" OR "#recklessdriving" OR "#assault" OR "#Pissing" OR "#spitting"'
searchQuery2 ='("Justin Bieber") ("throwing eggs" OR "drunk singing" OR "out of control" OR "found drunk" OR "spit on his fans" OR "struck a paparazzo" OR "shirtless photo" OR "found guilty" OR "reckless driving" OR "Pissing in buckets") '

# query='("Justin Bieber") ("throwing eggs" OR "drunk singing" OR "out of control" OR "found drunk" OR "spit on his fans" OR "struck a paparazzo" OR "shirtless photo" OR "found guilty" OR "reckless driving" OR "Pissing in buckets") '
rule=gen_rule_payload(searchQuery2, results_per_call=100)

# tweets = collect_results(rule, max_results=100, result_stream_args=premium_search_args) # change this if you need to

rs = ResultStream(rule_payload=rule,

tweets = list(

I get the following error:

retrying request; current status code: 403
retrying request; current status code: 403
retrying request; current status code: 403

Second, I tried using RESTful_Twitter_API and was able to collect 38 tweets.

client_key = '****'
client_secret = '****'

import base64,requests

key_secret = '{}:{}'.format(client_key, client_secret).encode('ascii')
b64_encoded_key = base64.b64encode(key_secret)
b64_encoded_key = b64_encoded_key.decode('ascii')

base_url = ''
auth_url = '{}oauth2/token'.format(base_url)

auth_headers = {
    'Authorization': 'Basic {}'.format(b64_encoded_key),
    'Content-Type': 'application/x-www-form-urlencoded;charset=UTF-8'

auth_data = {
    'grant_type': 'client_credentials'

auth_resp =, headers=auth_headers, data=auth_data)
# I got the accesstoken here.

search_headers = {
    'Authorization': 'Bearer {}'.format(auth_resp.json()['access_token'])    
search_params = {
    'q': '("Justin Bieber") ("throwing eggs" OR "drunk singing" OR "out of control" OR "found drunk" OR "spit on his fans" OR "struck a paparazzo" OR "shirtless photo" OR "found guilty" OR "reckless driving" OR "Pissing in buckets") ',
    'count': 100,
search_url = '{}1.1/search/tweets.json'.format(base_url)

search_resp = requests.get(search_url, headers=search_headers, params=search_params)

tweet_data = search_resp.json()

import pandas as pd
unsafe_data= pd.DataFrame(columns=['text', 'created_at'])
for x in tweet_data['statuses']:
    unsafe_data.loc[len(unsafe_data)]=[ x['text'],x['created_at']] 


Third, finally, I tried Tweepy python library which accesses Twitter Standard API.

import sys,os, jsonpickle, tweepy, simplejson as json

def data_maker(query, flag):
    '''1.It downloads tweets in *.json files at the tweets collected folder. It was necessary to bring all info. such 
        'created_at', user_id, etc. to see the behavior of Tweets' retrieval.
       2.Then we read these *.json files and create only *.txt files(containing only tweets) in order to run pre-processing script.
         It also outputs data frames to visualize in the notebook itself.
   # application's authentication
    auth = tweepy.AppAuthHandler(consumer_key="****",\
    #Setting up new api
    api = tweepy.API(auth, wait_on_rate_limit=True,wait_on_rate_limit_notify=True)
    maxTweets = 10000
    tweetsPerQry = 100
    count = 0
    data_dir = 'data_collection_methods/data/tweetsCollected/'
    if not os.path.isdir(data_dir):                                         
    file='%s.json' % flag
    #Open a text file to save the tweets to
    with open(file_path, 'w') as f:
        #keeping an assumption that we are only dealing with "English" tweets.
            for tweet in tweepy.Cursor(,q=query, lang='en').items(maxTweets) :         
                if tweet.text is not None:
                    f.write(jsonpickle.encode(tweet._json, unpicklable=False) + '\n')
                    count += 1
                    if count%1000==0:
                        print("Reached until",count,"and API call status is",api.rate_limit_status()['resources']['search'])

            print("Successfully grabbed", count, "tweets")
            print("file created but once the limit exceeds, API sends exceptions, I handled it.")
    # creating a df for the file
    df= pd.DataFrame(columns=['userId','text','created_at'])
    tweets_file = open(file_path, 'r')
    text_path   = os.path.join('data_collection_methods/data/tweetsText/','%s.txt' % flag) # create
    with open(text_path, 'w') as text:
        #first conversion into json object so that we can access it as a python Dict.
        for line in tweets_file:        
            tweet = json.loads(line)
            text.write(tweet['text'].replace('\n', '') + '\n') # removing any newline occurences & ADDing 1 in end and writing to .txt files.
            df.loc[len(df)]=[tweet['user']['id'],tweet['text'],tweet['created_at']]  # Just to visualize from/to dates & tweets in notebook .
    return df

#created queries and their respective dataframes calling above data_maker() and later after preprocessing merging similar dfs.
unsafeQuery1 ='"#JustinBieber" "#drunksinging" OR "#carelessDriving" OR "#outofcontrol" OR "#spitonfans" OR "#recklessdriving" OR "#assault" OR "#Pissing" OR "#spitting"'
unsafeQuery2 ='("Justin Bieber") ("throwing eggs" OR "drunk singing" OR "out of control" OR "found drunk" OR "spit on his fans" OR "struck a paparazzo" OR "shirtless photo" OR "found guilty" OR "reckless driving" OR "Pissing in buckets") '
safeQuery1   = '@justinbieber #fun OR #enjoy OR #Ecstasy'
safeQuery2   = '@justinbieber'

unsafe_df1=data_maker(unsafeQuery1, flag='unsafeQuery1')
unsafe_df2=data_maker(unsafeQuery2, flag='unsafeQuery2')
safe_df1=data_maker(safeQuery1, flag='safeQuery1')
safe_df2=data_maker(safeQuery2, flag='safeQuery2')
print('Initially :',unsafe_df1.shape, unsafe_df2.shape, safe_df1.shape, safe_df2.shape)

Result was:

Successfully grabed 0 tweets
Successfully grabed 38 tweets
Successfully grabed 3 tweets
Reached until 1000 and API call status is {'/search/tweets': {'limit': 450, 'remaining': 368, 'reset': 1520218776}}
Reached until 2000 and API call status is {'/search/tweets': {'limit': 450, 'remaining': 278, 'reset': 1520218776}}
Reached until 3000 and API call status is {'/search/tweets': {'limit': 450, 'remaining': 175, 'reset': 1520218776}}
Successfully grabed 3873 tweets
Initially : (0, 3) (38, 3) (3, 3) (3873, 3)

Kindly ignore the retrieval of 3873 tweets under “safeQuery2 = '@justinbieber'” query as I did this query to ensure there isn’t any problem with my code.

The problem is that if I try the above queries in any of the tweet retrieval processes, I am unsuccessful to grab the tweets for let say the year 2014-2017, but if I use the exact same queries and search in the “Search box”, I am able to see the results in that year. Could anyone help by?suggesting where I am going wrong and what I am doing wrong.


Hi @Anubhavfnu - do you have access to the premium Search Full-Archive beta? If you need data from 2012, you will need to use Full Archive. The searchtweets wrapper should work if you use your App credentials that you have set up for the dev environment existing in your portal here for the Full-Archive beta:


I am facing the same problem. I set my app environment under the archive/sandbox, and with 100 for both results_per_call and max_results in the program for a sandbox test first. But it provides me the error “HTTP Error code: 403: Forbidden: Authentication succeeded but account is not authorized to access this resource”. May I know what’s wrong or could you help me check my app or account access settings? my account is @michelle17058

By the way, it is usually we first test the program with sandbox and then goes to premium, right? And in the yaml credential file, if we are using sandbox, we can still write the account type as premium?

Thank you!