Closed lpsinger closed 3 years ago
@mcoughlin, I would like to unify the options for all three tiling methods (the geodesic tiling and the two spiral tilings). The geodesic tiling has one natural argument: the desired number of tiles. The reason that this is the natural argument is that it allows only certain integer values, and it must solve for the actual number of tiles given the requested minimum number of tiles.
The other two tilings take two arguments (fov, scale), but both of those arguments really control the average areal density of the tiles.
What would be the most natural argument(s) for all three tilings to take?
@lpsinger The "scale" term is redundant and a holdover from using the FOV in the config files. I do think something equivalent to "spacing between tiles" is what we are after in the end.
OK, then how about number of tiles, so that it's dimensionless?
My only concern is that people may not intuitively know how to calculate the number of tiles for a given FOV. Do we provide them some guidance maybe?
So maybe area per tile is better?
yes that sounds good to me.
The phi_theta_spiral
method is misnamed, isn't it? It's just a regular grid on a Lambert projection, that is to say, a regular grid in right ascension and cosine declination?
I misspoke. The phi_theta_spiral
is just a uniform grid on a sinusoidal projection, right, @mcoughlin?
Yes this is more accurate.
@mcoughlin, I've updated all of the methods to take an area parameter.
Please review both the golden angle spiral and the sinusoidal grid methods. I have reimplemented them to make the code more transparent.
Great! FYI I made some small changes to the indexing in the sinusoidal routine, to improve handling of the boundaries in ra and dec.
Refactor and clean up GROWTH tiling schemes