CIE doesn't like non-vismodel data

[thomased]

From someone trying to use MICA data:

fakedat <- data.frame(X = c(0.1, 0.2), 
                      Y = c(0.3, 0.4),
                      Z = c(0.5, 0.6),
                      lum = c(NA, NA))

colspace(fakedat, space = "cielab")

Error in h[h < 0] <- h[h < 0] + 360 : 
  NAs are not allowed in subscripted assignments

[thomased]

Ah, right. https://github.com/rmaia/pavo/blob/4903b8b9884c4e573ed8a4cdfa0ab4165dc80e55/R/cie.R#L65-L68

My head hasn't been in CIEspace for a while. I wonder if there's a possible workaround or sensible default we could 'guess' with a message....

[thomased]

One obvious idea would be to add illum and visual argument to cie() for those bringing their own data. Users could specify one of the built-ins if they happen to have used functions that match ours (which is possible for something like cie-10 degree & D65, for example). Or they can drop in the actual curves they used.

[Bisaloo]

Do we know if MICA will always use the same parameters to compute these values (e.g., always D65)? If so, it might be interesting at some point to add a function like import_mica() that sets all the attributes required downstream.

[Bisaloo]

This doesn't solve this issue though since users might import data from other sources :thinking:. I would go with https://github.com/rmaia/pavo/issues/201#issuecomment-641101693 as you suggest.

[thomased]

Yeah, seems like the simplest solution. MICA can in principle use anything for their calculations, like we can. I'm pretty sure we already share the same set of curves for viewers (cie-2 & cie-10) and illuminants (D65, forestshade etc.) though, which should make life easier for users.

Something like import_mica() is an interesting idea more generally though - worth some more thought.