goldingn

extend <- function (x, factor = 2) {
  # given an evenly-spaced vector `x` of cell centre locations, extend it to the
  # shortest possible vector that is at least `factor` times longer, has a 
  # length that is a power of 2 and nests the vector `x` approximately in the 
  # middle. This is used to define each dimension of a grid mapped on a torus
  # for which the original vector x is approximately a plane.
  
  # get cell number and width
  n <- length(x)
  width <- x[2] - x[1]

goldingn

# an 'unpack' operator, to do python/MatLab-like multiple assignment from
# functions

`%=%` <- function (expr, list) {
  
  if (!is.list(list))
    stop ('the right hand side must be a list')
  
  # grab the names to assign to
  expr <- substitute(expr)

goldingn

simulate_graf <- function (graf_model, n = 10, newdata = NULL) {
  # simulate n draws from the predictive posterior of a GRaF model, optionally
  # to new data
  
  # get the mean of the latent Gaussian process
  mean_f <- predict(graf_model,
                  newdata = newdata,
                  type = 'latent',
                  CI = NULL)[, 1]
  

goldingn

as_column_array <- function (x) {
  x <- as.array(x)
  if (length(dim(x)) == 1)
    dim(x) <- c(dim(x), 1)
  x
}

add_attribute <- function (object, attribute, name) {
  attr(object, name) <- attribute
  object

goldingn

library(greta)

# to be overwritten with correct method
greta_here <- function ()
  invisible(NULL)

# create an environment, and define the model there
greta_model <- function (model_expression,
                         parameters = list()) {
  env <- new.env()

goldingn

# devtools::install_github('goldingn/greta')
devtools::load_all()

fixed_values <- function (...) {

  # get the values and their names
  values <- list(...)
  names <- names(values)
  stopifnot(length(names) == length(values))

goldingn

# fake data
n <- 1000
m <- 50
x <- matrix(rnorm(n * m), n, m)
b <- rnorm(m, 0, 2) * rbinom(m, 1, 0.2)
eta <- x %*% b
y <- rnorm(n, eta, 0.3)

library (greta)

goldingn

devtools::install_github("RhoInc/CRANsearcher")
pkg <- CRANsearcher:::getPackages()

strings <- paste(pkg[, "Title"], pkg[, "Description"])

idx <- grep(" ecolog*| evolut*", strings, ignore.case = TRUE)

length(idx)
# [1] 236

goldingn

# install the experimental parallel branch
# remotes::install_github("zoonproject/zoon@parallel")
library (zoon)

# example workflow for 4 independent models that may take a while to run
run_wf <- function () {
  workflow(occurrence = UKAnophelesPlumbeus,
           covariate = UKBioclim,
           process = Replicate(Background(n = 1000), 4),
           model = GBM(max.trees = 10000),

goldingn

# quick & dirty caching of R objects - run the expression in b iff an RDS file
# for the object doesn't exist, otherwise load the object
`%<--%` <- function (a, b) {
  name <- deparse(substitute(a))
  file <- paste0(name, ".rds")
  if (file.exists(file)) {
    obj <- readRDS(file)
  } else {
    obj <- b
    saveRDS(obj, file)