PageRenderTime 10ms CodeModel.GetById 8ms app.highlight 0ms RepoModel.GetById 1ms app.codeStats 0ms

/jgogears/doc/TODO

http://jgogears.googlecode.com/
#! | 53 lines | 45 code | 8 blank | 0 comment | 0 complexity | a3389c192400723acefb94ae95d5dd23 MD5 | raw file
 1TODO
 2====
 3
 4Short Term
 5==========
 6* trace lack of symmetry in testTrainedModelEmptyBoard
 7* work on ASCIIToBoard -> BoardI round triping
 8* update javacc generated files in svn (eclipse javacc plugin is preventing this happening via eclipse)
 9* commandline program to play jgogears against gnugo
10* commandline program to guess moves from real games
11* decided whether to dual licence under the Apache 2 license
12* find a decent training corpus
13
14SGF Support
15===========
16* parse tt in SGF files
17* work out the differences between SGF 3 and 4
18* wider range of test inputs
19* find a better set of tests
20
21Board Support
22=============
23* 3-4 different types of KO detection
24* finish CachingKoRule
25
26General
27=======
28* Decent ASCII board output
29* Parse GnuGo board output
30  * direct
31  * cut and paste
32* Better test coverage
33* Silent test coverage
34* improve GTP support
35* SGF saving (use SGFC for validation)
36* generate and quality-check the javadoc
37
38Speed improvements, test and improve
39====================================
40* attribute parsing
41* game structure creation
42
43
44DONE
45* implement abstract KO detection 
46* fix java compiler style warnings (javadoc etc)
47* tests for ko rules
48* separate the Model/Node data structures from the algorithms
49* better tests of the Board implementations
50* jgogears.engine.Engine now up and running
51* make moves a little more random. They're meant to be random at the moment, but apparently aren't...
52* change training to only train on winning moves rather than all moves
53