Values 'n.def.' are not parsed correctly by the DatasetParser because InvariantCulture is assumed

[gkronber]

fixed with r405 and r406.

Since 'n. def.' should be treated as one token we can't blindly split tokens with white-space. Instead the new format uses '@' '=' ';' and '\t' to split tokens.

Meta data (variables) starts at the top of the file and continues while the first token in the line is a string. The data starts at the line where the first token can be parsed as a number. Exported files from HL2 are still compatible but the following is also legal:

PROBLEMNAME = XYZ
VARIABLENAMES = Date; X; Y; Z
TARGETVARIABLE = 3
NONINPUTVARIABLES = 3
MINIMUMTIMEOFFSET = 1; 1; 1; 0
MAXIMUMTIMEOFFSET = 6; 6; 6; 0
MAXIMUMTREEHEIGHT = 8
MAXIMUMTREESIZE = 50
TRAININGSAMPLESSTART = 13
TRAININGSAMPLESEND = 1513
VALIDATIONSAMPLESSTART = 1514
VALIDATIONSAMPLESEND = 2200
TESTSAMPLESSTART = 2201
TESTSAMPLESEND = 3709
19910221;6,5625;6,75;6,75;8,653;8,592;6,892;7,495;7,859;8,314;1565,59;2891,83;365,14;1,491;8,9125;9,02917;9,055
19910222;6,5625;6,75;6,75;8,666;8,605;6,971;7,569;7,898;8,322;1582,52;2889,36;364,97;1,5055;8,9175;9,03667;9,05417

[swagner]

Milestone 3.1 deleted

[swagner]

Milestone Iteration 0 deleted

[swagner]

Milestone Current deleted