Last active
October 20, 2022 22:27
-
-
Save macloo/5c69cdf5294fa97eb41d6ad950233cee to your computer and use it in GitHub Desktop.
Scrape members from a Twitter list and then get their bios, locations, etc.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
#!/usr/bin/env python | |
# encoding: utf-8 | |
""" | |
python 2.x | |
Scrape members from a Twitter list and then get their bios, locations, etc. | |
Based on | |
https://github.com/lamthuyvo/social-media-data-scripts/blob/master/scripts/twitter_bio_info_compiler.py | |
and | |
https://github.com/kylemanna/social-utils/blob/master/twitter/list-follow.py | |
""" | |
import tweepy # https://github.com/tweepy/tweepy | |
import csv | |
# import authentication credentials | |
from secrets import TWITTER_C_KEY, TWITTER_C_SECRET, TWITTER_A_KEY, TWITTER_A_SECRET | |
#authorize twitter, initialize tweepy | |
auth = tweepy.OAuthHandler(TWITTER_C_KEY, TWITTER_C_SECRET) | |
auth.set_access_token(TWITTER_A_KEY, TWITTER_A_SECRET) | |
api = tweepy.API(auth) | |
# which Twitter list and who owns it | |
slug = 'florida-legislators' | |
owner = 'JordanRaynor' | |
def get_list_members(api, owner, slug): | |
members = [] | |
# without this you only get the first 20 list members | |
for page in tweepy.Cursor(api.list_members, owner, slug).items(): | |
members.append(page) | |
# create a list containing all usernames | |
return [ m.screen_name for m in members ] | |
# create new CSV file and add column headings | |
def create_csv(filename, usernames): | |
csvfile = open(filename, 'w') | |
c = csv.writer(csvfile) | |
# write the header row for CSV file | |
c.writerow( [ "name", | |
"display_name", | |
"bio", | |
"followers_count", | |
"following_count", | |
"acct_created", | |
"location" ] ) | |
# add each member to the csv | |
for name in usernames: | |
user_info = get_userinfo(name) | |
c.writerow( user_info ) | |
# close and save the CSV | |
csvfile.close() | |
def get_userinfo(name): | |
# get all user data via a Tweepy API call | |
user = api.get_user(screen_name = name) | |
# create row data as a list | |
user_info = [ name.encode('utf-8'), | |
user.name.encode('utf-8'), | |
user.description.encode('utf-8'), | |
user.followers_count, | |
user.friends_count, | |
user.created_at, | |
user.location.encode('utf-8') ] | |
# send that one row back | |
return user_info | |
def main(): | |
# provide name for new CSV | |
filename = "userinfo.csv" | |
# create list of all members of the Twitter list | |
usernames = get_list_members(api, owner, slug) | |
# create new CSV and fill it | |
create_csv(filename, usernames) | |
# tell us how many we got | |
print "Number of rows should be %d, plus the header row." % len(usernames) | |
if __name__ == '__main__': | |
main() |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I imagine you'll have to iterate through your list and call sequence of functions on each item.