html_url,issue_url,id,node_id,user,created_at,updated_at,author_association,body,reactions,performed_via_github_app,issue
https://github.com/pydata/xarray/issues/3620#issuecomment-565638027,https://api.github.com/repos/pydata/xarray/issues/3620,565638027,MDEyOklzc3VlQ29tbWVudDU2NTYzODAyNw==,346587,2019-12-13T22:48:21Z,2019-12-13T22:48:21Z,NONE,"Thanks for the quick responses @rabernat and @dcherian!

> When you say “arbitrary complex mathematical functions”: what are the arguments / inputs to these functions?

The short answer is: pixels. Which in astro is sometimes the same as ""array index"", but usually off-by-0.5 (i.e., the *center* of the pixel), and occasionally off-by-1 (for files generated by e.g. FORTRAN or other 1-based indexing schemes...).

The longer answer is: it depends on which direction you mean.  Most WCS' are by design invertible, so you can go world-to-pixel *or* pixel-to-world.  In the former case that means there's a bunch of possibly inputs - examples include wavelength (for a spectrum), energy (for a photon-counting experiment), or altitude and azimuth (celestial coordinates that are similar to lat/lon but on the celestial sphere instead of a reference geoid).

Additionally, the WCSs have *parameters*, which are usually not thought of as ""inputs"" to the WCS, but rather internal state or metadata (usually they are stored in file headers and interpreted by software).

If you want to see some concrete examples, you might check out [Astropy's docs for our WCS interface](https://docs.astropy.org/en/stable/wcs/), or [the design document for the most recent iteration of that interface](https://github.com/astropy/astropy-APEs/blob/master/APE14.rst).  But those *might* be impenetrable for a non-astronomer, so I'm open to more follow-ups!

> It would also be good to hear about ""sub-pixel metadata""

Sorry I wasn't clear there - I meant there are cases where the sub-pixel structure of the coordinates matter - e.g. telescopes often have pretty intense optical distortion, and we sometimes interpolate in such a way that the sub-pixel information in the WCS is critical to getting the interpolation right.  And by ""metadata"" I meant both abstract parameters that encode that transformation (polynominal coefficients, information about which sky-projection is used, etc), and more physical metadata like ""the temperature of the observatory, which slightly changes the shape of the distortion"".  


> which we basically support through dask

Oh, tell me more about that - maybe this is the key for our needs?","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,537772490