在 Django 应用程序中嵌入 jupyter notebook/google colab

Question

I wanted to build a website and embed the jupyter notebook functionality of being able to create cells and run python code within it into my website

For creating a website I m using Django and I would like to embed either the google collab or jupyter notebook

By the way I have researched enough and have been stuck with the StackOverflow links where there no answer about this or the one where they want to use django in jupyter notebook

Thanks in advance for any guidance or any reference that you guys can provide.

Answer 1

You have many options to do that:

Note:: I used "jupyter-lab" you can use "jupyter notebook"

1- The first option to redirect to "jupyter notebook"

django view.py

from django.shortcuts import redirect,HttpResponse
import subprocess
import time

def open_jupiter_notbook(request):
    b= subprocess.check_output("jupyter-lab list".split()).decode('utf-8')
    if "9999" not in b:
        a=subprocess.Popen("jupyter-lab  --no-browser --port 9999".split())
    start_time = time.time()
    unreachable_time = 10
    while True and "9999" not in b:
        timer = time.time()
        elapsed_time = timer-start_time
        b= subprocess.check_output("jupyter-lab list".split()).decode('utf-8')
        if "9999" in b:
            break
        if elapsed_time > unreachable_time:
            return HttpResponse("Unreachable")
    path = b.split('\n')[1].split('::',1)[0]
    #You can here add data to your path if you want to open file or anything
    return redirect(path)

if you want to implement it in template instead of redirect, you can use the following code in Django template:

<iframe src="{% url 'open_jupiter_notbook' %}" width= 600px height=200px></iframe>

2- The second option:

just use jupyter notebook commands by using this subprocess.check_output("your command".split())

Answer 2

Export it first and then import it, look here for a good explanation

from google.colab import files
src = list(files.upload().values())[0]
open('mylib.py','wb').write(src)
import mylib

Answer 3

There is a medium article explaining how to do that here . The gist:

1. Install django_extensions
2. Add it to your INSTALLED_APPS
3. Run python manage.py shell_plus --notebook - you can run it in the background if you want jupiter to be accessible all the time.
4. You can redirect to the jupyter URL from django

Answer 4

A jupyter notebook is basically a kernel that runs (interactive) python or any other language on the backend and a web based frontend that communicates with this kernel. This communication is done by using zeromq. Under the hood it is websockets. The backend can run any other language as desired, eg Julia. The front end can be other things beyond web, like a terminal console or qt console (in this case however we would not call it jupyter notebook because the notebook part in the name stands for the web interface).

Server Side

In the python server side you will have to start the jupyter notebook server. Here we are using port 8123. We generate a random token that will have to be used in the client-side as well.

pip3 install notebook

from uuid import uuid4
from notebook.notebookapp import NotebookApp

PORT = 8123
TOKEN = str(uuid4())

app = NotebookApp()
app.initialize(["--port", f"{PORT}", f"--NotebookApp.token={TOKEN}"])

(response under construction - show server handler)

(full example in https://gist.github.com/bollwyvl/bd56b58ba0a078534272043327c52bd1 )

Client Side

All client-side behaviour you see in a Jupyter notebook is implemented in javascript. By client-side behaviour it is meant new cell creation, printing of results, showing graphs, establishing javascript side of communication et cetera. If you wish all of this then you would need almost

If you plan to have all of this behaviour then you would be better served by using an iframe including the regular jupyter notebook. If you plan having something simplier like having only a single cell and a full notebook is not necessary then you could implement a simple client by yourself. For intermediary solutions you could also take parts of the already done jupyter notebook page and take some parts of it for you (btw it uses BSD license).

Simple client side implemented by yourself

We are talking about client side, but since the client code has to be sent to client from server, still in the server you will have to instruct django to send the following html and javascript:

(response under construction)

Full third party client side

(response under construction)