Uploading tarballs some of the time is very confusing

[itamarst]

If there's more than one file to upload, it's a tarball. If it's one file, it's not a tarball.

This is very confusing as an end-user of this data.

dotscience-python/dotscience/__init__.py

Lines 445 to 463 in 9c84cdd

	tar.close()
	def _upload_output_files(self):
	outputFileSize = len(self.currentRun.metadata()["output"])
	if outputFileSize > 1:
	temp = tempfile.NamedTemporaryFile(delete=False)
	temp.close()
	# 1. Tar the files in the model dir
	self._tar_outputFiles(temp.name)
	# 2. Upload the tar
	# Uploading to the same model dir so we get proper paths
	# such as /model/assets/saved_model.json
	self._uploadArchive(temp.name, self.currentRun.getModelDir())
	os.remove(temp.name)
	elif outputFileSize == 1:
	self._upload(self.currentRun.metadata()["output"][0])
	print(".\n", end="")

The more reasonable thing to do would be to always upload a tarball.

[rusenask]

yeah, makes sense, the only issue is the way we upload/download them, tarballs where added as a way to support directories :|