Checklist for vegan 2.5-0 release

[jarioksa]

I wish to publish vegan 2.5-0 "asap". This "asap" may be as late as 4Q 2017, but with some luck it can be be earlier. There are big changes, and these are mega-issues that must be finished for the release:

• Finish new infrastructure for constrained ordination with several issues to be solved (detecting rank, handling zero-rank components). See issue common infrastructure for constrained ordination #218. This may be accomplished with commit 93da5ac, but must be checked.
• Fix regressions in constrained ordination with na.action = na.exclude (scores.cca() fails, some scores could be filled for excluded observations). Fixed by 439a905.
• Fix regression in rda that crashes analogue check. Fixed by 8efdf49.
• Check predict.rda, in particular for scaling: it seems that results change in analogue.
• Undocument cca.object: its internal structure is not in API and direct access (with $) should be discouraged. Instead we should describe accessor functions to internal structures. Done with commit db1a3ca, but we may need some new or updated accessor functions (issue cca support functions: something missing? #262).
• Finish moving to compiled code in anova.cca, potentially with new tests (issue Faster permutest.cca with compiled code #211).
• Registration and use of compiled code.
• Handle CRAN submission hassle with orditkplot. Closed via 28aeba1. (This was already released in 2.4-4 on Aug 24, 2017.)
• getF.c must check input (wrong dimensions in permutations can trash R). Closed via 22fe136.
• Decide what to do with capscale: deprecate or keep? See issue capscale: deprecate, downgrade or upgrade? #217.
• Decide what to do with new null models (issue nullmodel changes in 2.5-0 #255): there are two new boosted versions of quasiswap and at least one of them may be biased, and then there is new C-code w/o R-interface that is similar to r2dtable but with higher variance.
• Check CRAN check hassle with read.cep (issue Remove read.cep() from vegan? #263).

(This entry is the first version of this this issue and rather regarded as a place-holder for future checklist)

[gavinsimpson]

I know I haven't looked at the new MLM-based diagnostics yet (but now that summer is here, well spring on the prairies but classes ended anyway, I will look at those), but are you proposing that we merge these or is their absence from the list above indicative?

[jarioksa]

The MLM statistics were already merged to the master and as.mlm() functions are now deprecated (see issue #234). My main purpose was to list outstanding issues to remind what still must be done.