is there any related issue related to this enhancement?

@glemaitre No, I have not seen an issue related to this.

I'm working on comparing various non-linear dimensionality reduction algorithms and wanted to use the swiss roll and swiss roll with hole data. The swiss roll with hole is a classic dataset for evaluating non-linear dimensionality reduction and I felt that it's important to include the option to create it.

Could you please update/extend one of the existing examples to demonstrate how manifold learning models of scikit-learn behave on this dataset?

For instance, the following example could be extended into a multi-stage example (first without the hole and then with the hole):

https://scikit-learn.org/stable/auto_examples/manifold/plot_swissroll.html#sphx-glr-auto-examples-manifold-plot-swissroll-py

This example dates from a long time ago, before sphinx-gallery had support for notebook style examples. Nowadays we would write multi-step examples with the # %% cell separation marker. For instance look at for an example with a modern layout: https://scikit-learn.org/dev/auto_examples/applications/plot_cyclical_feature_engineering.html

@ogrisel Sure, I'll give it a try and update when done.

@ogrisel Sorry for the delay.

I have updated the example as requested and also included a comparison to t-SNE. Let me know if there's anything else I can do.