How to save requests (python) cookies to a file?

Question

How to use the library requests (in python) after a request

#!/usr/bin/env python
# -*- coding: utf-8 -*-
import requests
bot = requests.session()
bot.get('http://google.com')

to keep all the cookies in a file and then restore the cookies from a file.

Solution 1

There is no immediate way to do so, but it's not hard to do.

You can get a CookieJar object from the session with session.cookies, and use pickle to store it to a file.

A full example:

import requests, pickle
session = requests.session()
# Make some calls
with open('somefile', 'wb') as f:
    pickle.dump(session.cookies, f)

Loading is then:

session = requests.session()  # or an existing session

with open('somefile', 'rb') as f:
    session.cookies.update(pickle.load(f))

The requests library uses the requests.cookies.RequestsCookieJar() subclass, which explicitly supports pickling and a dict-like API. The RequestsCookieJar.update() method can be used to update an existing session cookie jar with the cookies loaded from the pickle file.

Solution 2

After a call such as r = requests.get(), r.cookies will return a RequestsCookieJar which you can directly pickle, i.e.

import pickle
def save_cookies(requests_cookiejar, filename):
    with open(filename, 'wb') as f:
        pickle.dump(requests_cookiejar, f)

def load_cookies(filename):
    with open(filename, 'rb') as f:
        return pickle.load(f)

#save cookies
r = requests.get(url)
save_cookies(r.cookies, filename)

#load cookies and do a request
requests.get(url, cookies=load_cookies(filename))

If you want to save your cookies in human-readable format, you have to do some work to extract the RequestsCookieJar to a LWPCookieJar.

import cookielib
def save_cookies_lwp(cookiejar, filename):
    lwp_cookiejar = cookielib.LWPCookieJar()
    for c in cookiejar:
        args = dict(vars(c).items())
        args['rest'] = args['_rest']
        del args['_rest']
        c = cookielib.Cookie(**args)
        lwp_cookiejar.set_cookie(c)
    lwp_cookiejar.save(filename, ignore_discard=True)

def load_cookies_from_lwp(filename):
    lwp_cookiejar = cookielib.LWPCookieJar()
    lwp_cookiejar.load(filename, ignore_discard=True)
    return lwp_cookiejar

#save human-readable
r = requests.get(url)
save_cookies_lwp(r.cookies, filename)

#you can pass a LWPCookieJar directly to requests
requests.get(url, cookies=load_cookies_from_lwp(filename))

Solution 3

I offer a way by json:

to save cookie -

import json
with open('cookie.txt', 'w') as f:
    json.dump(requests.utils.dict_from_cookiejar(bot.cookies), f)

and to load cookie -

import json
session = requests.session()  # or an existing session

with open('cookie.txt', 'r') as f:
    cookies = requests.utils.cookiejar_from_dict(json.load(f))
    session.cookies.update(cookies)

Solution 4

Expanding on @miracle2k's answer, requests Sessions are documented to work with any cookielib CookieJar. The LWPCookieJar (and MozillaCookieJar) can save and load their cookies to and from a file. Here is a complete code snippet which will save and load cookies for a requests session. The ignore_discard parameter is used to work with httpbin for the test, but you may not want to include it your in real code.

import os
from cookielib import LWPCookieJar

import requests


s = requests.Session()
s.cookies = LWPCookieJar('cookiejar')
if not os.path.exists('cookiejar'):
    # Create a new cookies file and set our Session's cookies
    print('setting cookies')
    s.cookies.save()
    r = s.get('http://httpbin.org/cookies/set?k1=v1&k2=v2')
else:
    # Load saved cookies from the file and use them in a request
    print('loading saved cookies')
    s.cookies.load(ignore_discard=True)
    r = s.get('http://httpbin.org/cookies')
print(r.text)
# Save the session's cookies back to the file
s.cookies.save(ignore_discard=True)

Solution 5

This will do the job:

session.cookies = LWPCookieJar('cookies.txt')

The CookieJar API requires you to call load() and save() manually though. If you do not care about the cookies.txt format, I have a ShelvedCookieJar implementation that will persist on change.

View more solutions

Comments

[Elmo]

requests.utils.dict_from_cookiejar and requests.utils.cookiejar_from_dict are not required. They don't save cookies with the same name for different domains and don't save all the required cookies data. I spent a lot of time debugging just because of these.

[ChaimG]

This answer is missing a few steps. Here is the full code: cj = cookielib.LWPCookieJar(cookie_file) cj.load() session.cookies = cj

[MattCochrane]

pickling a cookiejar doesn't seem to save the host that its associated with. when you load again all cookies are just in the host ''.

[MarSoft]

@MattClimbs, it is dict_from_cookiejar which doesn't save host information. Actually, in current version session.cookies can be pickled and unpickled directly, without converting to dict. Also, requests.utils.dict_from_cookiejar can be replaced with session.cookies.get_dict(), and cookiejar_from_dict can be replaced with session.cookies.update(my_dict).

[consatan]

all(isinstance(c, cookielib.Cookie) for c in r.cookies) is True, so why created each cookie item again in save_cookies_lwp?

[Javier C. H.]

using session.cookies.set(cookies.keys()[n],cookies.values()[n]) worked for me as well, thanks!

[User]

Please use a session in your answer

[Rontron]

This worked for me without the os.path.isdir lines in each function. Calling open() creates the file if it doesn't exist, and writes correctly for my use-case.

[Martijn Pieters]

Editors note: I've updated this top answer rather than add a new post. It was close enough but needed updating for API changes made in the intervening 6 years that make this task all that much easier. Do not use the dictionary utilities, they are not needed at all and do not preserve important cookie medatada. I'm fine with posting my own answer if the author wishes to revert my changes; I'd appreciate a heads-up in that case.

[Lennart Rolland]

requests.utils.dict_from_cookiejar <-- exactly what I wanted

[MarSoft]

Note that dict_from_cookiejar loses host information and hence does not work if there are several cookies with same name for different hosts.

[Emil]

This should have been best answer - no external libs which are not included by default.

[Jurass]

this is not correct, dict_from_cookiejar loses many informations about cookies.

[Weekend]

In my case (a python newbie), after open() as f, there should be a f.close(). Otherwise two consecutive open() would rise "TypeError: can't pickle _thread.RLock objects".

[user136036]

This basically works fine, but you cannot add/update cookies like you can with requests.update().

[GordonAitchJay]

@Weekend, you don't need to explicitly call f.close() when using the with statement. The context manager implicitly calls f.close() for you. See docs.python.org/3/tutorial/… The 2 separate with statements in the answer are not nested.