Support new KB API corpus changesCorpus KB API changes

[johnbradley]

Changes phenoscape_api() to return https://dev.phenoscape.org/api/v2-beta.
Fixes issue with anatomy_ontology_iris() returning PATO, and ZP.
Fixes errors with Resnik simlarity by removing sampling that occasionally caused problems.
Updates tests now that the KB API returns classification data in term info response.
Adds "states" to the corpus parameter for corpus_size() and term_freqs().

Fixes #243
Fixes #245
Fixes #246

[johnbradley]

I think there is a problem with term_freqs().
With the current changes there is an issue with passing an array of as values to term_freqs().
The /similarity/frequency endpoint has a new required value of type that only takes a single value.

rphenoscape/R/term-weights.R

Lines 54 to 59 in f2d05de

	term_freqs <- function(x,
	as = c("phenotype", "entity", "quality"),
	corpus = c("taxa", "taxon_annotations", "gene_annotations", "genes"),
	decodeIRI = FALSE,
	...) {
	as <- match.arg(as, several.ok = TRUE)

rphenoscape/R/term-weights.R

Lines 70 to 76 in 5ac0f17

	ontology_terms_type <- as
	if (ontology_terms_type == "entity") {
	ontology_terms_type <- "anatomical_entity"
	}
	query <- list(terms = as.character(jsonlite::toJSON(x)),
	corpus = corpus,
	type = ontology_terms_type)

[hlapp]

There are I think only two ways to fix this:

1. Change the as parameter to be a single value, and if in following the previous API a list is supplied, stop() if they're not all of the same value or if they are, change to that value.
2. Break the input list of terms into chunks corresponding to the same value in as. (This would only be needed if as is provided as a list of terms and they are not all the same.) Then process each chunk and finally re-assemble the list.

Having 2 would be nice. But one could consider 1 to be a first step in that direction by allowing as to be a vector, with the caveat that the values not all being the same is not yet supported.

[hlapp]

BTW is the following still true? I thought not?

rphenoscape/R/term-weights.R

Lines 66 to 67 in 5ac0f17

	if (any(as != "phenotype"))
	stop("corpus '", corpus, "' requires phenotype terms", call. = FALSE)

[johnbradley]

BTW is the following still true? I thought not?

Based on the new type parameter for/similarity/frequency that allows either phenotype and anatomical_entity I would assume not.

Blame says this commit was added via 330fc7d. I read through the referenced issues but am unsure what might have been the source of this code. Perhaps phenoscape/phenoscape-kb-services#146 (comment).

[hlapp]

@johnbradley are you still working on this? It looks ready to merge?

[johnbradley]

@hlapp I think the code is good now, but the documentation still needs work.

[johnbradley]

@hlapp Could you take a look at updating the documentation for term_freq() and corpus_size()?

[hlapp]

Looks all ready to go to me, and docs should now be good. Tests pass for me on local.

[johnbradley]

@hlapp Are you good with me merging the commits as they are or should I rebase/squash them?

[hlapp]

Let's leave the commits, I think most of them make sense on their own.