tests/testthat/test-SMG.Mean.R

In frbl/OnlineSuperLearner: Online SuperLearner package

context("SMG.Mean.R")
#==========================================================
described.class <- SMG.Mean 

context(" initialize")
#==========================================================
test_that("it should be able to initialize an objest with the correct attributes ", {
  colnames <- c('x1')
  subject <- described.class$new(colnames.to.mean = colnames)
  expect_that(subject,is_a('SMG.Mean'))
})

context(" process")
#==========================================================
test_that("it should return a dataframe with the correct names", {
  colnames <- c('x1', 'x2')
  subject <- described.class$new(colnames.to.mean = colnames)
  data <- data.table(x1=c(0,1,2,3),x2=c(5,2,1,0), y=c(1,0,1,0))
  result <-  subject$process(data)

  expect_equal(length(colnames(result)), length(colnames))
  expect_equal(colnames(result), c('x1_mean', 'x2_mean'))
})

test_that("it should return the correct means", {
  colnames <- c('x1', 'x2')
  subject <- described.class$new(colnames.to.mean = colnames)
  data <- data.table(x1=c(0,1,2,3),x2=c(5,2,1,0), y=c(1,0,1,0))
  result <-  subject$process(data)
  mean_x1 <- cumsum(data$x1) / seq(1, nrow(data))
  mean_x2 <- cumsum(data$x2) / seq(1, nrow(data))

  expect_true(all(unlist(result[,'x1_mean']) == mean_x1))
  expect_true(all(unlist(result[,'x2_mean']) == mean_x2))
})

test_that("it should be able to update an existing mean", {
  colnames <- c('x1', 'x2')
  subject <- described.class$new(colnames.to.mean = colnames)
  data <- data.table(x1=c(0,1,2,3),x2=c(5,2,1,0), y=c(1,0,1,0))
  subject$process(data)

  newdata <- data.table(x1=c(3),x2=c(5), y=c(1))
  result <-  subject$process(newdata)

  expect_true(all(unlist(result[,'x1_mean']) == mean(c(data$x1,  newdata$x1))))
  expect_true(all(unlist(result[,'x2_mean']) == mean(c(data$x2,  newdata$x2))))
})

context(" update")
#==========================================================
test_that("it should return the last means given the data", {
  colnames <- c('x1', 'x2')
  subject <- described.class$new(colnames.to.mean = colnames)
  data <- data.table(x1=c(0,1,2,3),x2=c(5,2,1,0), y=c(1,0,1,0))
  # Create an initial mean
  subject$process(data)

  newdata <- data.table(x1=c(3),x2=c(5), y=c(1))
  result <-  subject$update(newdata)

  print(result)
  expect_is(result, 'numeric')
  expect_true(all(unlist(result['x1_mean']) == mean(c(data$x1,  newdata$x1))))
  expect_true(all(unlist(result['x2_mean']) == mean(c(data$x2,  newdata$x2))))
})

context(" minimalObservations")
#==========================================================
test_that("it should return the minimal number of observations needed for a summary measure", {
  colnames <- c('x1')
  expected <- 1
  subject <- described.class$new(colnames.to.mean = colnames)
  expect_equal(subject$minimalObservations, expected)
})