Problem with coordinate implementation

[marina-ricci]

We recently merged a PR to specify the coordinate system for the shear, but it lacks some functionalities.
First, it should be explained in details in a notebook, so that this bug would have been found. E.g : showing a profile taken with an incorrect coordinate system and showing all intermediate steps.

At the moment, the keyword is missing from compute_tangential_and_cross_components. You can pass it to the Galaxy Cluster object, but it does not imply that it will be used (this should be avoided).

Also, we might consider adding a simple way to convert between the two coordinate, once compute_tangential_and_cross_components is run (not having to rerun all steps).

(Note : I discovered this while trying to implement in TXPipe and seeing no changes).

[caioolivv]

Hey @marina-ricci, just to be sure I understand what's the problem: GalaxyCluster.compute_tangential_and_cross_components() should have an option to pass in the coordinate system or should it explain that GalaxyCluster.coordinate_system will be used (or a combination of both)?

Also I agree there should be a function to convert between them. Maybe GalaxyCluster.convert_coordinate_system()? In that case, should you call it and then call GalaxyCluster.compute_tangential_and_cross_components() again or just calling GalaxyCluster.convert_coordinate_system() should already compute new tangential and cross components?

Finally, there's the examples/demo_coordinate_system_datasets.ipynb notebook from another PR which shows the problems of using another coordinate system. Is that what you meant? In that case we could reference it in another notebook?

[marina-ricci]

Hi @caioolivv ! I did not see that the examples/demo_coordinate_system_datasets.ipynb notebook was merged to main. I'm not sure it should stay here in it's current form.

These are my suggestions :

• add warning message when GalaxyCluster (or compute_tangential_xxx, see second point) object is initiated without specifying the coordinate system. Like "no coordinate system provided, taking euclidean as default"
• put coordinate system as a keyword for compute_tangential_and_cross_components, from data_ops or as a GalaxyCluster method. Maybe not necessary to have it as a GalaxyCluster keyword because it may be confusing as it is not necessarily propagated.
• create a simple notebook to highlight this.

I created a branch with a notebook to show the problem (it's not fully documented but should be understandable)