Complete restructuring

[xdvom03]

I may want to continue work on this. In particular, there are two possible ways forward:

a) Try to automatically create classes given data. This is tough (since there are many not-useful ways to class), but it might be doable using a good site-similarity measure and multidimensional tension analysis (spring-charge graph drawing, cluster analysis, that kind of thing). b) Use the classes I have and just plop them online as a web directory, then set the program up to fill it, publish the whole thing (including this repo) somewhere, see if anybody cares.

Both ways would be running more-or-less continuously in the optimal scenario.

However, before I move there, I must fix several presentation issues:

• The code must be divided into well-documented modules.
• The prefix hack should be removed. Learn ASDF @ configure Emacs to reopen stuff if needed (it shouldn't be).
• The UI must be remade - maybe dropping the grid in favour of place or pack (check how they work), or maybe another library if I get it working.
  • Unnecessary functions (like inputting text manually) should be removed or hidden.
• Errors must be handled well. The crawler cannot stop unless it is specifically meant to (i.e., the query has no solution).

[xdvom03]

On a return to the program, the GUI is clunky. A good way to begin: benchmark some actions simulated on a canvas, build ad-hoc UI with mouseovers 'n stuff if speed permits.

[xdvom03]

The biggest problem with the GUI is obvious enough: Too many damn buttons!! Everything is a button, so everything looks the same and you have to read tons of text. Here's all the stuff we can make not buttons:

• The listed pages (incl. the bucket)
• The classes
• The Move to classifier etc. buttons - could be a tab system.