Request's storage

[asvetlov]

Thinking on adding sessions to aiohttp.web I'we found that may be implemented as separate project.

API would be like:

from aiohttp_session import get_session

def handler(request):
    session = get_session(request)
    session['key'] = value

The session object may be implemented as Pyramid does (I just guess to borrow the code from aiorest: https://github.com/aio-libs/aiorest/blob/master/aiorest/session/base.py#L9).

Session loading and storing may be implemented by session middleware.

But I need the way to get session object from web-handler.

The most native way is storing session object in request by middleware on loading.

Those I guess to add request.storage property which is a regular python dict. Middleware may store the session in the dict under, say, 'aiohttp_session' key -- and user may return it back by get_session() call.

The same technique may be used, say, for saving database transaction for committing uncommitted transactions on web-handler finishing (and rolling back on exception from web-handler).

Another options is to save session as request._session (we use it for now but I don't like changing request's namespace, it should be considered as constant object).

The third option is to use global dict with Request objects as keys, but I hate singletons.

The fourth is to add Request.set_property() as Pyramid does but I personally don't like the idea of mutating the list of request public properties on request's execution. That may make a mess easy.

Resume: I propose to add per request dict storage (as well as we already have per application one) and allow to third-party libraries to make accessors to that storage if needed.

[fafhrd91]

how about Request(dict) same as Application?

[asvetlov]

I used to reject your idea before publishing the issue, but after rethinking I agree with you.

The same approach looks good.

[asvetlov]

Fixed in da0bd63

[ludovic-gasc]

You commit too quickly to have the time to react ;-)
Nevertheless, +2 with this proposal.

[asvetlov]

Sorry for that but I had have a feel that we are on the same point

On Wed, Jan 14, 2015 at 11:00 PM, Ludovic Gasc notifications@github.com
wrote:

You commit too quickly to have the time to react ;-)
Nevertheless, +2 with this proposal.

—
Reply to this email directly or view it on GitHub
#242 (comment).

Thanks,
Andrew Svetlov

[ludovic-gasc]

Don't worry, you're right ;-)

[lock]

This thread has been automatically locked since there has not been
any recent activity after it was closed. Please open a new issue for
related bugs.

If you feel like there's important points made in this discussion,
please include those exceprts into that new issue.