latest `fda` breaks `warpSet` function

[mikejiang]

flowStats failed to build on BioC lately due to the upgrade of fda package. Here is the call stack:

16: stop("WFDOBJ is not a single function")
15: monfn(x, Wfdobj)
14: fngrad.smooth.morph(y, x, wt, Wfdobj, lambda, Kmat, inact, basislist)
13: smooth.morph(xval, yval, WfdPar)
12: landmarkreg(fdobj, landmarks, WfdPar = WfdPar, monwrd = monwrd, 
        ...)
11: eval(expr, envir, enclos)
10: eval(expr, pf)
9: withVisible(eval(expr, pf))
8: evalVis(expr)
7: capture.output(regDens <- landmarkreg(fdobj, landmarks, WfdPar = WfdPar, 
       monwrd = monwrd, ...))
6: warpSet(x, parameters, ...) at #1
5: (function (x, parameters, ...) 
   warpSet(x, parameters, ...))(grouping = "GroupID", monwrd = TRUE, 
       x = <S4 object of class "flowSet">, parameters = c("CD8", 
       "CD69", "CD4", "HLADr"))
4: do.call(x@normFunction, args)
3: .local(data, x, ...)
2: normalize(Data(wf[["TCells+"]]), norm)
1: normalize(Data(wf[["TCells+"]]), norm)

[mikejiang]

By installing both versions of fda, stepping into warpSet and comparing the output from each fda function call, here we found the source of problem. The fda2.4.0::fdPar returns a single-column matrix

Browse[1]> WfdPar$fd$coefs
         [,1]
bspl4.1     0
bspl4.2     0
bspl4.3     0
bspl4.4     0
bspl4.5     0
bspl4.6     0
bspl4.7     0
bspl4.8     0
bspl4.9     0
bspl4.10    0
bspl4.11    0
bspl4.12    0
bspl4.13    0

yet the fda2.4.3::fdPar returns a n by n squared matrix

WfdPar$fd$coefs
         [,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10] [,11] [,12] [,13]
bspl4.1     0    0    0    0    0    0    0    0    0     0     0     0     0
bspl4.2     0    0    0    0    0    0    0    0    0     0     0     0     0
bspl4.3     0    0    0    0    0    0    0    0    0     0     0     0     0
bspl4.4     0    0    0    0    0    0    0    0    0     0     0     0     0
bspl4.5     0    0    0    0    0    0    0    0    0     0     0     0     0
bspl4.6     0    0    0    0    0    0    0    0    0     0     0     0     0
bspl4.7     0    0    0    0    0    0    0    0    0     0     0     0     0
bspl4.8     0    0    0    0    0    0    0    0    0     0     0     0     0
bspl4.9     0    0    0    0    0    0    0    0    0     0     0     0     0
bspl4.10    0    0    0    0    0    0    0    0    0     0     0     0     0
bspl4.11    0    0    0    0    0    0    0    0    0     0     0     0     0
bspl4.12    0    0    0    0    0    0    0    0    0     0     0     0     0
bspl4.13    0    0    0    0    0    0    0    0    0     0     0     0     0

This breaks fda::landmarkreg call later on.

diff on the code base of fda package further confirms this:

diff ../otherPkgs/fda_2.4.3/R/fdPar.R ../otherPkgs/fda_2.4.0/R/fdPar.R 
62c62
<         
---
>         coefs   <- matrix(0,nbasis-length(dropind),1)

For now we will simply inline the fdPar function from fda2.4.0 before its author addresses this issue.