Minimum Python Version

Pegasus Code except Worker Package

Lowest common version across supported platforms is

Worker Package

Lowest common version across supported platforms is  / 

Mats Rynge Can we expect Python 2.7 to be available on CentOS 6?

Mats Rynge List Python components included in worker package.

1. Use python-six compatibility library.
2. Refactor code such that Python 3 code is made Python 2 compatible instead of the opposite.
3. See: https://python-future.org/compatible_idioms.html
4. See: https://docs.python.org/3/howto/pyporting.html

Pegasus Code except Worker Package

Continue src/externals

Platform specific packages, python-flask

Will need to test with the lowest common available version across supported platforms.

Will require more platform specific integration tests.

Run pre/post install scripts to install using pip, i.e. pip install Flask.

Mats Rynge Do .rpm, .deb packages support this? Mats: this is only allowed if you do into your own space (which is similar to what we do with externals. You are not allowed to pip install to the system location)

Packages may conflict with system installed packages. yum install python-flask (v0.9) vs pip install Flask (v1.0.2)

Latter two will require more platform specific integration tests to detect incompatible, missing dependencies.

Worker Package

Continue src/externals

Only for packages not expected to be available on supported platforms.

"Vendorize"

Copy dependency code into Pegasus.vendor directory.

# Instead of this
import six

# Use this
import Pegasus.vendor.six

Testing

Environment

Will run unit tests on the minimal supported Python version and above.

Using pytest instead of unittest

Uses assert instead of self.assert*.

import pytest

# Simple method, no need for classes
def test_method():
  # Simple assert, no need for assertEqual, etc.
  assert a % 2 == 0, "value was odd, should be even"

  with pytest.raises(ZeroDivisionError):
        1 / 0

# One method, multiple tests
# https://docs.pytest.org/en/latest/parametrize.html
@pytest.mark.parametrize(
    "a, b",
    [
		(1, 2, 3),
		(2, 3, 5),
		(5, -100, -95),
	],
)
def test_eval(a, b, expected):
    assert a + b == expected

# Dependency Injection
@pytest.fixture(scope="function OR class OR module OR package OR session")
def client():
	import requests
 	s = requests.Session()
	s.get(".../login")
	yield return s # Yield can be replaced with return if no cleanup is required.
	s.get(".../logout")
    s.close()

def test_external(client):
  assert client.get("/endpoint-1")

See: https://docs.pytest.org/en/latest/contents.html