Maximizing VariableImpactView shows no data

[fholzing]

Reproduce as follows: Create Random Forest Regression for a reasonable large dataset (Tower). Open the Variable Impacts in the Random Forest Regression Solution. The data will show. Now double-click on the Variable Impacts entry in the Regression Solution. The View will maximize, but no data will be displayed.

[fholzing]

r17276: Changed cancelation-initiation from VisibleChanged to OnHidden and added a flag, preventing an error in case the View gets closed.

[pfleck]

Tested r17276 and it works.

However, I also tried a simpler solution to the problem, by only triggering the cancellation token within RegressionSolutionVariableImpactsView_VisibleChanged (line 88) if this.Visible is false. Could you test if this solution has any drawbacks that I am not thinking of?

[pfleck]

@fholzing, I attached a patch that hopefully fixes the progress disposing issue. Please check if you can still produce any exceptions when applying the patch (to the stable).

[pfleck]

r17426

• Certain events of the TargetControl are now forwarded to the ProgressView in order to correctly dispose and hide the ProgressView.
• Simplified hiding the ProgressView within the Regression/ClassificationSolutionVariableImpactsView.

I could not reproduce any of the tickets issues. However, I still managed to get a "long lasting" progress bar for when optimizing a symbolic expression tree within the simplifier. I will look further into this, why the progress bar is sometimes lingering that long, even after cancelling the operation.

[pfleck]

r17430

• Fixed progress calculation for InteractiveSymbolicDataAnalysisSolutionSimplifierView.
• Added a minimum improvement (of 1E-10) to stop constant opt.
• Added tooltips for Simplify and Optimize.

Concerning reducing the repetitions for constant opt to 100 (instead of 1000): In #2844, it was stated that it should not be necessary to click the optimize button multiple times, thus the high repetition count. However, if we want to show a meaningful progressbar, having such a high number of repetitions would lead to a very slow progress that will prematurely close anyway due to convergence of the algorithm. After discussion with mkommenda, we opted for reducing the number of repetitions in favor of a reasonable progress indication.

[fholzing]

I retested both the initial issue concerning the missing data when maximizing the VariableImpactView and the new behaviour of the ConstOpt/ProgressBar. No issues occured, neither for Regression- nor Classification-solutions. Seems to work like a charm.

Just one very minor remark: I highly approve the tooltips! Maybe the simplify tooltip should also be added to the SymbolicClassificationSolutionView and SymbolicRegressionSolutionView (btnSimplify) instead of the generic "Simplify solution" tooltip?

[mkommend]

r17488: Updated tooltip for simplify button in all symbolic data analysis solution views.

[mkommend]

Reviewed r17276, r17426, r17430, r17488.

Merge problems might occur in connection with #3042.

[mkommend]

r17500: Merged r17276, r17426, r17430, r17488 into stable.