Rperp metric and max_rpar for 3-point statistics

[caspervedder]

Hi Mike,

First of all, thank you for making your code public!

I was wondering if it would be possible to have some sort of 'projected/transverse' metric which could be used in combination with a restriction on the line-of-sight separation of the points for the 3-point correlation functions (in my case, GGG specifically). This would be similar to the Rperp metric for 2-point correlation functions (discussed in Issue #17).

I guess this can be made quite complicated as there are 3 parallel coordinates for the three points of the triangle. However, a simple restriction would already be very helpful for me. For example, if we order the parallel distances: r_par1 < r_par2 < r_par3, a helpful restriction would be |rpar3 - rpar1| < Pi.

In general, I can imagine there might be several other options that could be relevant for different purposes though, specifically if the cross-correlations NNG and NGG are added.

Best,

Casper

[rmjarvis]

The Rperp metric should already been enabled for 3pt calculations. But you're correct that it's not currently possible to put restrictions on r_par combinations. I'll think about how to implement this.

[caspervedder]

Thank you! As for the Rperp metric, I seem to be missing something here. If I make a very simple GGGcorrelation() object with metric = 'Rperp', I get the following error:

This seems to match the documentation (https://rmjarvis.github.io/TreeCorr/_build/html/correlation3.html):

Edit: if I add the metric in the process function, like ggg.process(cat, metric = 'Rperp'), it does not give this error. (although it does return only 0's, but this could very well be due to other reasons, it also gives a failed assert output.)

[rmjarvis]

Oh, you're right. Sorry, I forgot that 3pt limited to that. I think it should be easy to enable all the metrics for 3pt. I can't remember now if there was a good reason that I limited to those three, but I don't think it should be a fundamental issue. The rpar limits will be a bit more work though.