Try to get replies to a particular set of tweets, recursively.

replies.py

#!/usr/bin/env python

"""

Twitter's API doesn't allow you to get replies to a particular tweet. Strange
but true. But you can use Twitter's Search API to search for tweets that are
directed at a particular user, and then search through the results to see if 
any are replies to a given tweet. You probably are also interested in the
replies to any replies as well, so the process is recursive. The big caveat 
here is that the search API only returns results for the last 7 days. So 
you'll want to run this sooner rather than later.

replies.py will read a line oriented JSON file of tweets and look for replies
using the above heuristic. Any replies that are discovered will be written as
line oriented JSON to stdout:

    ./replies.py tweets.json > replies.json

It also writes a log to replies.log if you are curious what it is doing...which
can be handy since it will sleep for periods of time to work within the 
Twitter API quotas.

PS. you'll need to:

    pip install python-twitter

and then set the following environment variables for it to work:

  - CONSUMER_KEY
  - CONSUMER_SECRET
  - ACCESS_TOKEN
  - ACCESS_TOKEN_SECRET

"""

import sys
import json
import time
import logging
import twitter
import urllib.parse

from os import environ as e

t = twitter.Api(
    consumer_key=e["CONSUMER_KEY"],
    consumer_secret=e["CONSUMER_SECRET"],
    access_token_key=e["ACCESS_TOKEN"],
    access_token_secret=e["ACCESS_TOKEN_SECRET"],
    sleep_on_rate_limit=True
)

def tweet_url(t):
    return "https://twitter.com/%s/status/%s" % (t.user.screen_name, t.id)

def get_tweets(filename):
    for line in open(filename):
        yield twitter.Status.NewFromJsonDict(json.loads(line))

def get_replies(tweet):
    user = tweet.user.screen_name
    tweet_id = tweet.id
    max_id = None
    logging.info("looking for replies to: %s" % tweet_url(tweet))
    while True:
        q = urllib.parse.urlencode({"q": "to:%s" % user})
        try:
            replies = t.GetSearch(raw_query=q, since_id=tweet_id, max_id=max_id, count=100)
        except twitter.error.TwitterError as e:
            logging.error("caught twitter api error: %s", e)
            time.sleep(60)
            continue
        for reply in replies:
            logging.info("examining: %s" % tweet_url(reply))
            if reply.in_reply_to_status_id == tweet_id:
                logging.info("found reply: %s" % tweet_url(reply))
                yield reply
                # recursive magic to also get the replies to this reply
                for reply_to_reply in get_replies(reply):
                    yield reply_to_reply
            max_id = reply.id
        if len(replies) != 100:
            break

if __name__ == "__main__":
    logging.basicConfig(filename="replies.log", level=logging.INFO)
    tweets_file = sys.argv[1]
    for tweet in get_tweets(tweets_file):
        for reply in get_replies(tweet):
            print(reply.AsJsonString())

Could you give a example for tweets_file?
I wonder what should be written in the file

@Allen-Qiu If anyone is wondering the file should be in jsonl (as the code describes) You get that file from twac or tweepy (those are the format outputted by library)

The Twitter API v2 supports this now using a conversation_id field. You can read more in the docs.

First, request the conversation_id field of the tweet.

https://api.twitter.com/2/tweets?ids=1225917697675886593&tweet.fields=conversation_id

Second, then search tweets using the conversation_id as the query.

https://api.twitter.com/2/tweets/search/recent?query=conversation_id:1225912275971657728

This is a minimal example, so you should add other fields as you need to the URL.

Absolutely, v2 is the way to go now! We have support in twarc for doing it too:

$ twarc2 conversation 1225912275971657728 > tweets.jsonl

or, if you have a file of tweet ids:

$ twarc2 conversations ids.txt > tweets.jsonl