stitch.rspec function for row-wise merging

[thomased]

Function for stitching (row-merging) rspec objects of differing wavelength ranges. It has functionality for handling specs with partly-overlapping wl regions (averaging, min, max), as well as non-overlapping wl regions (via interpolation). Can handle rspec objects when the spec names match in each but are in a different order, or when they contain only some common samples (with the other being NA filled for the relevant wl region). Edit: whoops, didn't realise it'd contain the entire commit history for #247 too.

• Addresses create stitch.rspec for row-wise merges #175.
• Implement generic/method
• Add tests
• Update news
• Any other use-cases it should handle?
• Update handbook (post-merge)

[Bisaloo]

Function name? I'm still slightly confused on the whole method/generic thing in this case. Ideally it'd be stitch.rspec().

Here is the general gist:

• If a generic is already defined for this in your package or one of the dependencies (including base), you just have to provide the new method as you did. This applies to function such as mean(), plot(), summary(), etc.
• If a generic is not defined anywhere (this is your case in this PR), you have to provide the generic and the method

#' @export
stitch <- function(x1, x2, ...) {
  UseMethod()
}

#' @export
stitch.rspec <- function(x1, x2, ...) {

}

Success can be verified by the presence of a S3method() line in NAMESPACE. E.g.,

pavo/NAMESPACE

Line 12 in b1144fd

S3method(subset,colspace)

[Bisaloo]

I am proposing some changes here: stitch...stitch-2 to reduce special handling of edge cases. As far as I can tell, everything can be dealt with through the general case. If you agree with them, I can push them to this branch.

The implementation looks good to me. I have left one comment on design regarding argument names and whether we want to use S3 or not

[Bisaloo]

Suggested change

	stitch <- function(rspec1, rspec2, overlap_method, interp) {
	UseMethod("stitch")
	}
	#' @rdname stitch
	#'
	#' @export
	stitch.rspec <- function(rspec1, rspec2,
	overlap_method = c("mean", "minimum", "maximum"),
	interp = TRUE) {
	stitch <- function(x1, x2, ...) {
	UseMethod("stitch")
	}
	#' @rdname stitch
	#'
	#' @export
	stitch.rspec <- function(x1, x2,
	overlap_method = c("mean", "minimum", "maximum"),
	interp = TRUE) {

To leave room for future extensions or other methods.

We may want to re-use this generic for other objects in the future and it's probably safer to offer more flexibility in the argument names. This is the same reason why we use x as the first argument in subset.rspec(), plot.rspec(), etc.

If we don't expect ever re-using this generic for other objects, we could also not leverage S3 and propose a stitch_rspec() or stitchrspec() function directly. The same way we do it with explorespec(), aggspec(), etc.