Multi-Scatterplot in Preprocessing crashes HL with high number of variables

[pfleck]

Because the number of charts in the multi-scatterplot scales quadratic with the number of variables, a high number of variables causes HL to crash (limitation of system-handles).

To limit the displayed variables, a list of checkable variables should be displayed, such as in the linechart and histogram. If the number of pre-checked variables is exceeding a certain threshold (on opening the chart first time), the user should be asked if he really wishes to display all variables, risking HL to crash, or wishes to only pre-check a limited number of variables.

Furthermore, the list of checkable variables should take the allowed input variables of the problem data into account; for visual highlighting and functionality to only check allowed input variables and the (optional) target variable.

[pfleck]

r14381

• Refactored CheckedVariablesView out of the ChartView to allow reuse of the checked variables list.
  • The new list visualizes the non-input/target variables in gray.
  • Added context menu to quickly (un)check all variables or only the inputs+target variables.
• In the Multi-Scatterplot
  • New structure and layout of the single charts to support fixed header rows and columns (for the variable names). Instead, removed the legend of each plot for better usage of plot area.
  • Adapted the new CheckedVariablesView (but hidden until (un)checking is implemented).

[pfleck]

r14382 Forgot to uncomment abstract keyword.

[pfleck]

r14384 Implemented unchecking of variables in multi-scatterplot.

[pfleck]

r14388

• Implemented (re)checking of variables in multi-scatterplot.
• Added 2 sliders for changing the chart size.

[pfleck]

r14418 Only input and target variables are pre-checked for linechart, histrogram and scatterplot.

[pfleck]

r14425 Fixed wrongly formated variable-entries in checked item list.

[gkronber]

Please include all files (PreprocessingCheckedVariablesView.cs?) in the project.

[pfleck]

I did not find any non-included files. PreprocessingCheckedVariablesView.cs is already included in the project.

[mkommend]

Reviewed and tested the functionality of the ScatterPlotMultiView.

Review comments

• Maximum of 20 variables should be selected by default.
• All controls should be displayed (variable check box, slider, ...) before the charts are drawn (asynchronously?).
• One slider for the chart size is sufficient. It is quite cumbersome to handle two sliders. The point size should also be adapted when the chart size changes.
• Possible leak (memory, window handles). Removed controls are not disposed.
• Inserting and removing charts is quite ugly and pretty slow. Another possibility would be to set the column / row width to 0. Maybe that is a "better" solution.

[gkronber]

The improved multi-scatterplot is nice!

[pfleck]

Because the Preprocessing Enhancements branch (#2709) was created after the changes of this ticket, I suggest we release this ticket and implement the review comments in #2709 to avoid potential merge conflicts.

[gkronber]

r14963: merged r14381 r14382 r14384 r14388 r14418 r14425 from trunk to stable

[gkronber]

r14917 (see comment:11) has not been merged as this is seemingly tracking in another ticket (#2709).