attic/uploadOpenMLRun.R

In openml/r: Open Machine Learning and Open Data Platform

#' Upload an OpenML run to the server.
#'
#' Share a run of an implementation on a given OpenML task.
#'
#' @param task [\code{\link{OMLTask}}]\cr
#'   The task.
#' @param mlr.lrn [\code{\link[mlr]{Learner}}]\cr
#'   The mlr learner, if used. Optional.
#' @param impl.id [\code{\link{character}}]\cr
#'   The ID of the OpenML implementation that is stored on the OpenML server.
#' @param predictions [\code{data.frame}]\cr
#'   The predictions. Must have the same form as produced by \code{\link{runTask}}. If a fatal error
#'   occured while computing predictions, you have the possibility to leave this out and pass an error
#'   message instead (as parameter \code{error.msg}).
#' @param error.msg [\code{character(1)}]\cr
#'   An optional error message. If this is not empty, the predictions (if present) are ignored.
#' @param session.hash [\code{character(1)}]\cr
#'   A session token returned by \code{\link{authenticateUser}}.
#' @param run.pars [\code{list}]\cr
#'   If no mlr learner was used, this is a list of all \code{\link{OpenMLRunParameter}s}.
#' @param clean.up [\code{logical(1)}]\cr
#'   Should intermediate files be removed from disk at the end?
#'   Default is \code{TRUE}.
#' @template arg_showinfo
#' @return [\code{numeric(1)} or \code{NULL}]. Run ID if the run was uploaded succesfully.
#' @export

uploadOpenMLRun = function(task, mlr.lrn, impl.id, predictions, error.msg, session.hash,
  run.pars = NULL, clean.up = TRUE, show.info = getOpenMLOption("show.info")) {

  if (!missing(mlr.lrn)) {
    assertClass(mlr.lrn, "Learner")
    run.pars = makeRunParameterList(mlr.lrn)
  }
  assertList(run.pars)
  assertClass(task, "OMLTask")
  impl.id = asCount(impl.id)

  if (missing(error.msg)) {
    assertDataFrame(predictions)
  } else {
    assertString(error.msg)
  }

  run.desc = makeOpenMLRun(
    task.id = task$id,
    implementation.id = impl.id,
    parameter.settings = run.pars
  )

  if (!missing(error.msg))
    run.desc$error.message = error.msg

  description = tempfile()
  on.exit(if (clean.up) unlink(description))
  writeOpenMLRunXML(run.desc, description)

  file = tempfile()
  on.exit(if (clean.up) unlink(file), add = TRUE)
  if (show.info) {
    messagef("Uploading run to server.")
    messagef("Downloading response to: %s", file)
  }

  url = getAPIURL("openml.run.upload")

  if (!missing(predictions)) {
    output = tempfile()
    on.exit(if (clean.up) unlink(output), add = TRUE)
    write.arff(predictions, file = output)

    content = postForm(url,
      description = fileUpload(filename = description),
      predictions = fileUpload(filename = output),
      session_hash = session.hash
    )
  } else {
    content = postForm(url,
      description = fileUpload(filename = description),
      session_hash = session.hash
    )
  }
  write(content, file = file)
  # was uploading successful?
  doc = try(parseXMLResponse(file, "Uploading run", "upload_run"), silent = TRUE)
  # if not, print the error.
  if (is.error(doc)) {
    parseXMLResponse(file, "Uploading run", "response")
  }

  if (show.info)
    messagef("Run successfully uploaded.")

  return(if (!is.error(doc)) xmlRValI(doc, "/oml:upload_run/oml:run_id"))
}