I completely rewrote the templater, using more of a parsing technique. (I still wish I could've gotten Syntax to work for Python, but just having a JFlex grammar is not enough.) The big obvious result is that it actually works; you can see it by going here with login prr:arf. (Use "all the news from the most recent scan"--I think I broke the others.) If the item got the default time for having no or an unparseable date, the time is not displayed. If the item has a comments tag, the note Qbullet is displayed. (I didn't realize those tiny iconic UserLand images images were all Qbullets: that globe, the magnifying glass, the orange "?" help button, everything.)
Biggest current problem: HTTP trouble. bIPlog from the Berkeley journalism school was unreachable last night, but it was some kind of "busy unreachable" that kept the browser open, waiting and waiting. Prrarf waited too, so it just sat there, waiting for the resource that wasn't coming. I need to see if there's a socket timeout option I can set (all the way through httpcache and urllib2, heh).
The tiny add.cgi I whipped up works OK, but I should have it use rssparser to make sure it's an RSS feed and get the feed title. (I also made it require POST so you can't use it with the prr:arf login, but then I realized the bookmarklet can't POST either. So I need to make it spit out a form to submit to itself when GETted instead of that 400 error.) I used the bookmarklet to add Erik Benson's feed to my list, and the bookmarklet grabbed the wrong feed (for he links two, one for the weblog and one for the Morale-O-Meter). I didn't realize it until reading the aggregator because the CGI simply says "Feed added a-OK!" Also, the Morale-O-Meter feed doesn't look terribly nice, with a description of just "+" or "-".
I set up a couple personal Roundup trackers on McGroarty's server so I can remember project ideas and issues. Or rather, so I don't have to remember them. I found that the current releases (which aren't what I was running for the existing Roundup tracker I have) have a "topic" field, which is perfect for tracking issues in multiple personal projects in one tracker--except only in the bleeding-edge version will topics group properly (topic is a Multilink field, which is what it sounds like: a sequence of references to instances of another type, in this case type keyword). I'm using the packaged Debian version, as McGroarty loves all over the Debian; I'm not sure if I'll upgrade to 0.6b1 or try to backport the fix. McGroarty suggested the latter and it'd certainly be easier than setting up a third version of Roundup. (I think we also have an earlier version installed already, for the other tracker, making that three. Ugh.) That's only related to Prrarf in that it's one of the projects I'm tracking there.
Also I got my changes to httpcache.py sent back to Joe Gregorio, which is nice.
