Changelog v2.0.1

Changes from v2.0.0 to v2.0.1

Release Date: January 29, 2009

This is a bug fix release, which addresses the following issues:

  • Fix: PMML reader did not retain ordering of input columns (bug #1592)
  • Fix: "Cluster Assigner" node assigned wrong cluster due to randomization of input columns (bug #1592)
  • Fix: PMML Cluster reader did not fully read cluster model; ignored last normalization element (no case number)
  • Fix: "PMML Reader" threw unhandled exception when underlying file was not available (bug #1633)
  • Fix: PMML Regression reader was not strict enough about potential normalization parameters; it will now fail during execution if normalization parameters are encountered (no case number)
  • Fix: SVM Learner did not write correct parameters into PMML prediction model (bug #1655)
  • Fix: Weka predictor had problems when test data class column has fewer instances than training data (bug #1641)
  • Fix: All Weka Learner nodes did not report errors properly (turned green but did not provide model) (bug #1657)
  • Fix: Rule Engine partially replaced settings with default values (bug #1593)
  • Fix: "Rule Engine" choked on missing values in the input table (bug #1594)
  • Fix: "Rule Engine" node did not accept "MISSING" as starting expression (bug #1595)
  • Fix: All tables were kept in main memory if old workflow was converted into v2.0 format (bug #1631)
  • Fix: "HTML Report" node did not include images (bug #1597)
  • Fix: "Math Formula" did not handle missing values for min and max function (bug #1635)
  • Fix: Error message, when OS browser widget is not found, was too long (on Linux, no case number)
  • Fix: NullPointerException in default column component for column filtering when last spec is null (bug #1639)
  • Fix: Using Concatenate node in workflow loops issued confusing error message (bug #1645)
  • Fix: Workflow loop variables were visible outside the loop (no case number)
  • Fix: Fuzzy rule learner node did not output full spec when domain contained more class labels than the rule set covered (no case number)
  • Fix: Loop end nodes did not complain if input specs of loop iterations differed (bug #1650)
  • Fix: "Many 2 One" node couldn't execute if input table already contains certain columns (bug #1647)
  • Fix: Node Views were not unregistered from model when node was canceled (no case number)
  • Fix: Decision Tree Learner had problems with numeric columns (ignoring some of the attributes) (bug #1654)
  • Fix: Plotter framework did not paint axis labels correctly (bug #1666, #1667)
  • Fix: NodeFactory did not resolve image paths with only two characters (no case number)
  • Fix: DataBase Reader nodes did not handle all DB types correctly, specifically strings (no case number)
  • Fix: BatchExecutor did not interpret "-workflowFile" argument if zip contained workflow in separate folder (bug #1597)
  • Fix: Some older Linux distributions (e.g. CentOS 3.9) did not resolve the OS browser widget correctly (no case number)
  • Eclipse 3.4 compliant (basic tests only) (no case number)