suppliedElsewhere is ignoring module loadOrder in some cases

[ianmseddy]

In suppliedElsewhere there is a for loop which iterates over other module's depenencies. Because del is either filtered or re-assigned to dd within each iteration, the only load order that is being considered is that of the final module in otherModDeps.

inFutureInit <- if (any(c("i", "c") %in% forms$where)) {
    del <- depsEdgeList(sim, plot = FALSE)
    # if ("c" %in% forms$where) {
      # The next line is subtle -- it must be provided by another module, previously loaded (thus in the depsEdgeList),
      #   but that does not need it itself. If it needed it itself, then it would have loaded it already in the simList
      #   which is checked in a different test of suppliedElsewhere -- i.e., "sim"
      dd <- del[objName %in% objDeparsed][from != to][!(from %in% c("_INPUT_")), ]
      d <- depends(sim)
      otherModsDeps <- d@dependencies[which(!names(d@dependencies) %in% currentModule(sim))]

      for (mod in otherModsDeps) {
        lo <- mod@loadOrder
        if (!is.null(lo$after)){
          del <- dd[from %in% lo$after]
        } else del <- dd #THIS LINE IS CAUSING THE ISSUE
      }

      # if (any(c("i", "c") %in% forms$where)) {
      # The next line is subtle -- it must be provided by another module, previously loaded (thus in the depsEdgeList),
      #   but that does not need it itself. If it needed it itself, then it would have loaded it already in the simList
      #   which is checked in a different test of suppliedElsewhere -- i.e., "sim"
      # if (exists("aaaa", envir = .GlobalEnv)) browser()
      out <- del[!(from %in% c("_INPUT_", currentModule(sim))), ][
        objName %in% objDeparsed]
      out <- out[, .(objName, noFeedback = all(from != to)), by = from][noFeedback %in% TRUE]
      objDeparsed %in% out$objName

Reprex (the tricky part was getting box and AI to install correctly)
The error will be an geographic extent error if Biomass_core supplies the object (because it is at the extent of rasterToMatch, not rasterToMatchLarge), or it will be an error that speciesLayers is missing in the Init of BBDP if Biomass_core does not supply it. In this case, it appears to depend on loadOrder and whether Biomass_core's loadOrder metadata is updated to include after = c("Biomass_borealDataPrep").

repos <- c("https://predictiveecology.r-universe.dev", getOption("repos"))
source("https://raw.githubusercontent.com/PredictiveEcology/pemisc/refs/heads/development/R/getOrUpdatePkg.R")
getOrUpdatePkg(c("Require", "SpaDES.project"), c("1.0.1.9003", "0.1.1.9009")) # only install/update if required



projPath = file.path(getwd(), "suppliedElsewhere_reprex")
Require::Require("PredictiveEcology/reproducible@AI")
Require::Require("PredictiveEcology/SpaDES.core@box")

out <- SpaDES.project::setupProject(
  paths = list(projectPath =  projPath),
  modules = c("PredictiveEcology/Biomass_borealDataPrep@development",
              "PredictiveEcology/Biomass_core@main",
              "PredictiveEcology/Biomass_regeneration@development" #need some extra module to trigger bug
  ),
  options = list("reproducible.inputPaths" = "~/data"),
  params = list(),
  packages = c('RCurl', 'XML', 'snow', 'googledrive', 'httr2', "gert", "remotes",
               "PredictiveEcology/LandR@development"),
  times = list(start = 2011, end = 2012),
  studyArea = {
    sa <- LandR::randomStudyArea(size = 10000*6.25*5000)
  },
  studyAreaLarge = {
    terra::buffer(studyArea, 2000)
  },
  rasterToMatchLarge = {
    rtml<- terra::rast(studyAreaLarge, res = c(250,250))
    rtml[] <- 1
    rtml <- terra::mask(rtml, studyAreaLarge)
  },
  rasterToMatch = {
    rtm <- reproducible::postProcess(rasterToMatchLarge, cropTo = studyArea, maskTo = studyArea)
  },
  sppEquiv = {
    speciesInStudy <- LandR::speciesInStudyArea(studyAreaLarge, dPath = "inputs")
    species <- LandR::equivalentName(speciesInStudy$speciesList, df = LandR::sppEquivalencies_CA, "LandR")
    sppEquiv <- LandR::sppEquivalencies_CA[LandR %in% species]
    sppEquiv <- sppEquiv[grep("_spp", KNN, invert = TRUE)]
  },
  loadOrder = unlist(modules)
)

outSim <- SpaDES.core::simInitAndSpades2(out)

[eliotmcintire]

I have pushed a change that is now being tested. Branch: suppliedElsewhere