Daniel's Blog

Weekend Program: Lisp Tumblog

Posted by Daniel Lyons Mon, 26 Nov 2007 18:16:00 GMT

Yesterday, I wrote a 250 line Lisp program to run my new tumblog. I have a few notes here I wanted to share about it. The next version will probably be open source; right now I can hardly stand to look at the code.

1. Hunchentoot is very cool. The prefix and regex dispatcher functions are very handy. It was actually very easy to use, so easy it was almost counter-intuitive. I wonder why it took me so long to start actually building this program, since I’ve been thinking about it for so long.

   However, it doesn’t help you build an abstraction atop it. That’s your problem. So my URL handling is very ad-hoc and slimy, which is one reason pagination doesn’t work right now.

2. CL-WHO is a mixed bag. On the one hand, it seems to be very efficient. On the other hand, it’s a bit hard to figure out how to build a template with it. It also can’t handle XML namespaces, which means my Atom feeds won’t validate due to missing XML namespace declarations and whatnot. I also don’t know how to make it emit an XML DOCTYPE or the XML processing instruction which indicates the encoding. Once I figure out how to do a reasonable templating system with CL-WHO my line count will probably go down by 50 lines or so.
3. cl-prevalence was a pretty serious win in the storage backend department. I just make a list of instances and as long as I go through the transaction mechanism they’re saved to disk and reappear when I restart the app. Very simple and elegant. I wish I hadn’t had to go by the really old documentation to figure it out though.

   There is one bug with cl-prevalence though, which is that it doesn’t handle the clock—or at least it doesn’t seem to based on what I could find. This is a problem because my instances save their instantiation time as their creation time. I worked around it by making a snapshot immediately after adding an instance to the list. I have mixed feelings about that. In any case, it shouldn’t be very hard to fix this in the library by providing a special variable with the time when the transaction started and telling people to use that instead of (get-universal-time).

4. I included the Parenscript library but never used it. So it’s Ajax-free for right now.
5. Initially I was very worried about what format to put the date in. Then I decided, fuck it, nobody really cares what time you post something, they only care whether or not the site is abandoned or if they’ve missed something. So I put the last-updated date at the bottom and my design is even more minimal than Steve Dekorte’s. Top that! :) (There’s no way it could be related to my being a shitty designer.)
6. Developing with Slime and sshfs was pretty good, but sshfs was definitely the worst part of it. I realized last night I could have just had the code checked out via Mercurial locally and be sending that to the remote Lisp, and then eventually check it in and back out on the server so that startup would work. Kind of a “duh” moment.
7. There was a general lack of refactoring. I have a to do list with about 5 simple refactorings I should have done during development. Once a little bit of your code is scuzzy, it’s pretty easy to let the rest of it turn scuzzy apparently. I guess programming is no different than any other aspect of life.
8. Weirdly, when I call read-from-string in Slime, it comes back in the tumblog: namespace, but when my code for processing the form calls it, the atoms come back in the common-lisp-user: namespace. It’s moderately annoying because I’m having to use #'string-equal to compare atoms instead of #'eq.

So there you have it. A short feature review: I can post through the web (but not edit), it understands markdown for my commentary, you can filter by tag, it has Atom feeds overall and by tag, and that’s it. I will hopefully rewrite it soon to be less scuzzy, or perhaps to add pagination, after all I only have 4 more entries to post before I need it. I plan on using tags to point out items of particular interest to some of my friends and family, so if you see your name on an item, that’s why.