Good patience to all of you. And great work to MenTaL, who's been
fielding patches and doing good things. Same for William and Frederick.
Tonight I've checked in my last week of work. I've been kind of
indecisive, leaving my Hobix utterly unstable for the interim. But I
think things are progressing well, but expect some growing pains here,
there's still alot of work to do.
But if you have time to help and these new changes excite you: hey, come
on, do it.
== Properties ==
The biggest changes of late have been related to handling properties
within Hobix. You'll notice that the Hobix::Weblog object now uses
these methods to.
The result is:
class Weblog
include BaseProperties
_! 'Basic Information'
_ :title, :req => true, :edit_as => :text
_ :link, :req => true, :edit_as => :text
_ :tagline, :req => true, :edit_as => :text
_ :copyright, :edit_as => :text
_ :period, :edit_as => :text
_ :lastn, :edit_as => :text
_! 'Entry Customization'
_ :entry_class, :edit_as => :text
_ :central_prefix, :edit_as => :text
_ :central_ext, :edit_as => :text
_! 'Paths'
_ :entry_path, :edit_as => :text
_ :lib_path, :edit_as => :text
_ :skel_path, :edit_as => :text
_ :output_path, :edit_as => :text
_! 'Participants'
_ :authors, :req => true, :edit_as => :map
_ :contributors, :edit_as => :map
_! 'Links'
_ :linklist, :edit_as => :omap
_! 'Sections'
_ :sections, :edit_as => :map
_! 'Libraries and Plugins'
_ :requires, :req => true, :edit_as => :omap
end
What you see above are section headers followed by properties. It's
simple sugar for the same old stuff you've already seen in Hobix, it's
just no longer tucked away in the property_map methods.
== Bayesian Filtering ==
The new bayesian classification is a part of the search engine.
Overall, it's pretty slow (the entire database is stored in a couplet of
Marshal dumps), but you can hook the bayesian engine to store all kinds
of information to rate content. It could potentially be used to rank
comments Slashdot-style or to flag comments you like as :important and
search for them later on.
The filtering itself doesn't work yet for comments, I'm still deciding
where to put the spam comments. I'm thinking of hanging on to them in a
corpus index in case the search engine ever needs to be regenerated.
The search engine is currently rather slow, but I'm hoping to move to
Hobix::DataMarsh, which is a DBM which stores Marshal records. At least
it won't incur loading of the complete database.
== The Publisher ==
This is probably something many of you have been dying to get your hands
on. I'm afraid it's still quite primitive, but you can give it a try by
using the patch or by simply adding 'hobix/publisher' to your requires
list, given that you have the CGI stuff installed.
== A Few Anticipations ==
1. Contributors: I want to get rid of the 'contributors' accessor for
the Weblog. It's supposed to work like 'authors' does, in that you can
assign these contributors to posts and track their contributions -- how
many posts did I yank off Anarchaia, how many posts are flgr
contributions?? -- but it doesn't work at present and I just don't have
time to make it work.
2. Bayesian + Tags: I think I want to change bayesian filtering to use
tags directly. This would give tags indexing and would allow you to use
tags as enumerated lists.
In other words, if you have a :spam and a :ham tag, you can ask the
database, do the words in this comment lean towards the :spam tag or the
:ham tag? Likewise, you could supply [:important, :personal, :todo,
:later] as a list and ask the database to scale all posts along that
spectrum, assuming you view them as a sort of priority scale.
3. Spam Blacklists: I personally don't like blacklists. I'm into
filters. But do any of you want a blacklist plugin?
Yawn, yawn, scratch, scratch.
_why
_______________________________________________
Hobix-is-the-way mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/hobix-is-the-way
