Getting Bryar to work: A Summary
Posted by matijs 11/10/2004 at 12h40
As I mentioned, installing Bryar was not
completely straightforward. Luckily, it is architected well, so making the
necessary modifications was easy.
First, comments didn’t work. Clicking the comments link simply
returned the same page. After some digging through the templates, and then
through the Template Toolkit documentation to find out how templates
worked, it turned out the toolkit was unable to access bryar.arguments
(where the CGI parameters live). As a result, the template toolkit couldn’t
access the CGI parameters, so it couldn’t know it was supposed to show the
comments.
I subclassed
Bryar::Renderer::TT to pass the parameters as a separate
variable to the toolkit. I still don’t know why the original didn’t
work. [Update: Now I know. It is a known bug.]
Putting the date inside the entry’s file worked (using
Bryar::DataSource::FlatFile::Dated), but only if I entered it in seconds since the epoch. So, I wrote a different subclass of Bryar::DataSource::FlatFile, one that parses the first line as a date
using Data::Parse.
The blog now worked, but it was slow: it took two seconds on the 333 MHz
Celeron laptop that is now my blog server. To fix this, I made a port of the
mod_perl frontend to mod_perl2 (I use Apache 2). I also modified it
somewhat: there is now a separate BryarDriver class that simply
calls Bryar->go(), essentially replacing bryar.cgi. This makes things a
lot cleaner, I think. [Update: Since the mod_perl2 API went through some changes, there is now a version for mod_perl2 before version 1.999.23 and a version for later versions of mod_perl2.]
Finally, the atom feed was broken: the code that filled the summary
field didn’t take into account that it’s possible for HTML tags to be
spread out over multiple lines. Luckily, this was easily fixed by modifying
the template to first collapse each entry to a single line.