Work with the back end to provide a way for algorithms to be transferred to the back end (in the form of trees) and retrieve the results after the data is retrieved and algorithms run on it.
First, DESIGN the rating tree structure. This tree should satisfy the following:
-The tree itself should fully define any ranking algorithm we need, parsable using Visitors (look at the extended visitor pattern, which is in the Common TFS folder and used by ocntent, I believe).
-Should allow for sorting "in order", that is, sort first by criteria A, then by criteria B. For example, sort first by ranking title (student, teacher, expert) and then by creation date.
-Should allow for sorting using weights, that is, combining different algorithms using weights. For example, sort with personal trust at weight 10 and global trust at weight 1. Then, local weight would be 10 times more important than global.
Then,
-Write up a detailed specification of the design and send it to Aaron/the backend team. This should include a diagram and descriptions of each class, much like the
selection tree spec did. Post it under Project documents, architecture documents, search and navigation should see "trees.one" there, which is the selection tree spec...i need to rename that)
(DUE BY MONDAY)
-Replace the only ranking tree files with the new structure.
Should be little to no "real code", as the structure should do most of the definitions.
I see the tree as having two complex types, "ComplexWeightedRankingTree" and "ComplexOrderedRankingTree". Then there is the "AbstractSimpleRankingTree" which has a "type" field (which like in the selection tree, tells the backend what table/column to search/sort by). There would be two concrete implementations, one saying the order is greatest to least, the other least to most. I feel the tanking tree can only go upon something with some sort of "numerical" value, such a a value of trust, a creation date, etc. This is what I see, but if you see room for improvement, by all means, have at.