writePackage fails with models

[paciorek]

  code <- quote({
        for(i in 1:5) 
            for(j in 1:3)
                y[i,j] ~ dnorm(0,1)
    })
    data <- list(y = matrix(rnorm(15),5))
    mclass <- nimbleModel(code, data = data)
writePackage(mclass,dir='/tmp',pkgName='foo',modify='clear')

Error in parse(file, keep.source = keep.source) : 
  /tmp/foo/R/mclass.R:4:53: unexpected '<'
3: 61502494123, -1.8686242436452476, -1.095839078217715, 0.58871129992911397, 0.68521054275100535, 0.61673494409436969, 0.6602703613553077, -1.1899422113694018, -0.22289784115920333, -1.006918771
4: 0.91911081127748462), dim = c(5L, 3L))), modelDef = <

Just reporting for the moment.

[perrydv]

This appears to be happening because (in reduced form) deparse(list(A = new.env())) returns list(A = <environment>). In mclass, modelDef and dataRules$y both deparse as <environment>. So clearly the scheme of deparsing the class generator and writing it to file for package code is not really general. This probably needs some discussion and thinking. Does it block you in any work at the moment or can you proceed with non-package compilation?

Could we just save the nClass (R6) generator as an object in the generated package?

[paciorek]

no, I am not blocked. I ran into it while debugging so it was a bit inconvenient but then I figured out the bug.