Skip navigation

Category Archives: Something New Each Month

For February I tinkered with Django, arguably the most popular web framework for Python. Prior to this, the only web framework I’ve had experience with is CodeIgniter and one proprietary JSP-based framework.

In an earlier post I covered installing Django. Nothing much has changed between then and now except that the most recent version of Django, 1.5, now supports Python 3! The first for Django ever. But I’m sticking with Python 2 for the time being.

Note: I tried learning Django last August, in an afterthought project with goals loftier than what my resources could have achieved. Needless to say, I never got too far with it. It was this which prompted me to specify a mini-project I’d use to learn web frameworks (any web framework). I’m not sharing the specifics of it now but you can gather what you can in my sandbox repository.

The Admin Site

One nifty feature of Django is how it creates an admin site automatically for you. On my CodeIgniter projects, I’d pattern the users table something like this, borrowing from Linux file systems:

CREATE TABLE IF NOT EXISTS users(
    -- Some project-specific fields here
    create_time TIMESTAMP NOT NULL,
    last_updater INTEGER NOT NULL,
    last_update TIMESTAMP NOT NULL,
    can_read BOOLEAN NOT NULL DEFAULT TRUE,
    can_write BOOLEAN NOT NULL DEFAULT TRUE,
    can_exec BOOLEAN NOT NULL DEFAULT FALSE,
    PRIMARY KEY (userid),
    FOREIGN KEY (last_updater) REFERENCES users (userid)
    -- Yes, apparently, MySQL allows you to foreign key recursively
);

If I’m really feeling up to it, or if the project requires (never had it tho), I might even implement user groups. But even with this, I can already define 23=8 types of users, which is already way more than what one project usually requires. What I’d normally do here is to designate can_exec privileges to admins. Way neater than maintaining different tables for different user types and then giving admins backdoor log-in pages—one of the naive approaches I used when I first started designing databases for apps.

Django automatically covers this for you, user permissions, groups, and all. The structure becomes a bit different though, and I find permission management way more complicated than my binary-scheme. Take a look at all possible user permissions in Django:

django_permissions

Look at how many that is, and realize that there is still a scroll bar!

What Django basically did is to allow you to specify add/change/delete permissions for every table in your database. Note that due to how Django enforces code modularity (see below), the number of tables in your database could easily blow-up.

Code Modularity

Perhaps one of the most confusing things I encountered as I tried to learn Django was how it distinguishes the terms “project” and “app”. Normally, I’d talk about an “app” being a “project” for some party. But in Django a project is a collection of apps. I had a hard time designing my project around these terminologies even after reading discussions at StackOverflow until I realized that perhaps a better term for “app” would’ve been “plug-in” and there’s really nothing stopping you from making your project as one big plug-in.

With the side-effect, of course, of your app being non modular and probably not well-factored. So, yes, the way Django handles things enforces the nice practice of modularity. Nice to take advantage of it, though it may be too time-taking to wrap your design around Django’s rules.

Django also enforces a form of Object-Relational Mapping (ORM). In fact, you don’t need to write a single line of actual SQL code to get your database up and running—all DB creation is done through the ORM and manage.py . This even allows you to easily switch DB back-ends. Then again, as with above, not all projects would fit well for an ORM approach.

Overall?

I find Django pretty neat given that you only need minimal set-up for this whole bunch of admin stuff. But then I haven’t had much time to explore it although I’ve volunteered in a (non-work, non-paid) project that might use Django pretty extensively. Guess I’ll get more Django this month. Let’s see.

Okay. I know it’s already February and that it’s kinda late. But it’s in times like this that I say “better late than never!”. But as I’ll show, I’ve already started with my resolutions—I’m just writing about them now.

But since this is my coding blog, the resolution I will mention here will be my technical-life resolution (as opposed to the personal ones everybody makes). So, without further ado…

I resolve to learn a new technology every month. Or, at least, get deeper with one I’m already acquainted with.

Also, I’d go into programming competitions/contests again. Team requirements limit my choices but, hey, there are a lot I can do alone online, not to mention free! (Though if anyone wants to team-up with me, let’s talk.)

And, just so I’ll stick with it, I’ll write a blog post every month about what I’ve been up to. Consider this the one for January.

For my first resolution, I created this sandbox repository at GitHub. And since my GitHub account is, in itself, already my sandbox of sorts, I guess this is some sandbox-ception eh?

January, I did sockets in Python. I didn’t manage to dig as deep as I’d have wanted because the Facebook Hacker Cup kicked in earlier than I expected and that falls under my second resolution. So it had to give way. Besides, I’m already dealing with a few sockets too many at work.

Regarding my second resolution, I don’t really expect much from it. At least, I won’t get rusty from the lack of problem sets in the real world (read: industry). At most, I might score a free trip to some world finals. But that world-finals scenario is still very out of my league. I’ve been doing ACM-ICPC problems since I was in my sophomore year in College and even now I find most of them really tough and tricky. Also, after less than a year out of school, I’m bound to have some rust in my system (not that school is a sure fire way to keep you sharp though). I only decided to start practicing with my algorithms sometime in December so there isn’t really much to expect.

(Note: I started writing this post in the last hour before the first round of Hacker Cup started. It’s now over and everything fell within my expectations. There’s a lot to learn. Maybe, I’ll write about this soon.)


So, what do I think about sockets in Python? (This will be short since, as mentioned, I didn’t have much time to dig deep into this.)

As with all things Python, it looks super neat. I got a client and server up and running in just 41 lines of code combined. The biggest savings comes from the fact that socket I/O is direct in Python; you no longer have to create OutputStream and InputStream objects as the socket objects themselves have methods for sending and receiving. It’s also interesting to note that Python’s socketserver class has a serve_forever method which, as the name implies, handles requests as long as it can (the official phrase is: “until an explicit shutdown signal is received”).

Oddly, this design reminds me how I handled one of the things I did at work. Won’t go into the details but I thought that it’s neat to have one class responsible for managing one whole session: packet numbers, windowing, etc. One instance of a class maps to one client; when client goes away so does this instance (at least, I no longer care if the garbage collector marks it as garbage). Makes OO live more peacefully with concurrency.

Awesome that Python enforces this design by default. Really, the more I discover about Python, the lovelier programming gets.

‘Till I get my February experiments done! ~Chad