Pickling Django request objects

Question

I'm trying to pickle a request object so that I can run my view code in a debugger with valid input. However, I'm getting

Can't pickle 'lock' object: <thread.lock object at 0x93ad240>

I looked through the request object, but couldn't find a thread.lock object anywhere in it. Does anyone know where it is? Is there a better way to go about this?

Answer 1

Generally, you cannot pickle HTTP requests objects for any web server. Pickling process does not pickle on the object itself, but all its references.

The usual web server design pattern is to have each HTTP request running its own thread. There must be a connection between the thread and HTTP request. Thus, the HTTP object is tied to the life cycle of web server and cannot be taken out of the web server context.

What you probably want to do is to pickle the contents of HTTP request object and re-create requests using this content as a payload.

Answer 2

Ok, for those who are interested, I have managed to pickle HttpRequest and WSGIRequest objects. First, you will need to monkey-patch at least one class, WSGIRequest (and probably HttpRequest), so that it exposes a __reduce__(self) method. You would something like:

WSGIRequest.__reduce__= __reduce__

The reduce method may look like this:

def __reduce__(self):
    meta = dict([(k,self.META[k]) for k in METACOPY if k in self.META and isinstance(self.META[k], str)])
    return (HttpRequest, (), {'META':meta, 'POST':self.POST, 'GET':self.GET, 'user':self.user, 'path':self.path})

where METACOPY is a list of keys you want to keep, eg. ['REMOTE_ADDR']

I find it handier and more transparent than the payload method (which I used with celery before).