Re: New session module

Answering your questions:

> Hi!
>
> No time to try this unfortunately but I definately agree with all your
> points.  I remember trying to use sessions for the first time and I
> almost gave up before finding session.auto.  Even with that, I do not
> like the setup required.  So, if your module does do what you say,
> then I'd vote for it replacing the current session module.  Then
> again, maybe I am missing something others can chime in on?

It does do what I said, but I'm also open to suggestions if people  
don't like my implementation.  I tested it considerably, but I've  
been known to throw a bug in my code once or twice in my life.  It  
may have a lurking bug somewhere. :)

>
> Also... concerning my last post, will your session module deal with
> non pickle able objects or at least throw an error?

The in-memory method (web server installation only) can store  
anything, pickable or not.  But the default method, dbm, can only  
take picklable objects.  There really isn't a way around this.  It's  
a fact that occurs when you try to save something to disk from any  
language, not just python.

However, I could throw a nice error explaining the situation to an  
unknowing user.  I'll do this.

>
> Cheers.
> Jon.
>
> On 12/28/05, Conan C. Albrecht <conan-RXmsyT/Szrj2fBVCVOL8/[email protected]> wrote:
>> Spyce developers (Rimon & Jonathan in particular):
>>
>> Attached to this email is a new session module, called session2.  I'd
>> like feedback on it, and if it passes muster after people comment and
>> I modify it appropriately, I'd like to either replace the existing
>> session module with it or add it as a second alternative in the main
>> source tree.
>>
>> I've never been happy with the existing session module.  I think it's
>> a weakness of Spyce right now.   I hope I'm not the only user who
>> feels this way.  (If I am, I'll go back to my corner :)  The
>> session.auto is not obvious, and there are too many options.  Most
>> users just want to save data for a given user in a cache and get it
>> on subsequent requests.  The existing module supports this, but it
>> requires too much setup and doesn't have the "batteries included"
>> feel.  I want a session module that is seamless and automatic,
>> similar to sessions in competing platforms.
>>
>> This new session2 module is easy and automatic.  In short, it acts
>> like a dictionary.  It can store arbitrary (hashable) items just like
>> a regular dictionary can. The contents of the "dictionary" change
>> automatically depending upon the user (browser cookie) is viewing the
>> page.  The programmer doesn't have to worry about cookies, disk
>> files, or anything else.  He/she just puts stuff in the dictionary
>> and gets it on subsequent calls.
>>
>> Following is example code using the new module:
>>
>> [[.import name=session2]]
>> [[\
>>    session2['username'] = 'Bart Simpson'
>>    print session2.get('username')        # prints Bart Simpson
>>    print session2['username']            # prints Bart Simpson
>>
>>    del session2['username']              # deletes the key/value from
>> the session
>>    #print session2['username']           # throws a KeyError (so
>> commented out)
>>    print session2.get('username')        # prints None
>>    print session2.get('username', 'Me')  # prints Me since it was the
>> default
>>
>>    session2['visited'] = session2.get('visited', 0) + 1
>>    print 'You visited us %s times.' % session2['visited']
>> ]]
>>
>> I've tried to program it as efficiently as possible.  Sessions get
>> created just-in-time, so if you never put anything in a person's
>> session (meaning you just retrieve from it), you get the default
>> values without any memory or disk access happening.  It spreads
>> sessions across disk files in a clever way, providing efficient
>> access.   It is also thread-safe.
>>
>> It defaults to saving session data in dbm-files since I estimate most
>> installations of spyce are mod_python, cgi, or default fastcgi (i.e.
>> multiple instances of spyce).  A memory-cache is also available for
>> those using web-server installations.
>>
>> It only supports cookies right now.  I haven't supported passing the
>> sessionid via the request because it requires URL rewriting for every
>> URL on the site (which is a lot of work).  I believe cookies are
>> widely accepted on the net now -- you can't access many sites with
>> cookies disabled anymore.  The point of this module is to be seamless
>> and automatic, and URL rewriting is not seamless or automatic. :)  If
>> people want this, I can add it, but I think it kind of defeats the
>> point of the module.
>>
>> The only real problem is it is not process-safe.  If two instances of
>> spyce hit the same dbm file on disk at the same time, they might
>> blast the same file at the same time.  The existing session module
>> doesn't prevent this either.  I'm not sure it it is possible to
>> prevent this in an efficient way.
>>
>> So... I present this module for your review.  Drop it in your modules
>> directory and try it.  Comments are welcome.
>>
>> Conan
>> ____________________________________
>> Conan C. Albrecht, Ph.D.
>> Information Systems Department
>> Brigham Young University
>>
>> Web: http://warp.byu.edu/
>> Email: conan-RXmsyT/Szrj2fBVCVOL8/[email protected]
>> Skype: doconix
>> (please IM in active working groups; send email otherwise)
>>
>>
>>
>>
>>
>>
>>
>>