Testing
Headless tests, GUI tests and screenshots.
Everything runs Blender with a throwaway user profile, so your own preferences, extensions and generated add-ons are never touched. Blender comes from $BLENDER or BLENDER_EXECUTABLE in config.yaml.
python scripts/test.py # headless suite (tests/test_*.py)
python scripts/test.py -k core -v # filter by test id, verbose
python scripts/test.py --gui # also tests/gui/*.py in a real window (undo, ...)
python scripts/screenshot.py tests/visual/scenarios/live_panel.py out.pngWhat is covered
test_nodes_compile.pyadds every node, chains it into program and interface flows and as a data source, then checks that the generated add-on compiles and loads. A new node is covered without writing anything.test_core.py: the pipeline. Hot reload, rollback on errors, renames, undo, duplicated trees, references across trees, export.test_nodes_*.py: focused checks of generated code per node family.test_performance.py: a 300 node chain must stay fast.
Writing a test
Tests run inside Blender with unittest. Use tests/helpers.py instead of reaching into the add-on:
import unittest
import helpers
class ListTest(unittest.TestCase):
def setUp(self):
helpers.reset_file()
def test_for_each(self):
tree = helpers.new_tree()
trigger = helpers.add_node(tree, "SNA_Node_Trigger")
loop = helpers.add_node(tree, "SNA_Node_ForEachList")
helpers.link(tree, trigger.outputs[0], loop.inputs[0])
helpers.flush() # run the rebuild now instead of on the timer
self.assertIn("for index_1, item_1 in enumerate(", helpers.tree_source(tree))Screenshots
tests/visual/scenarios/*.py and website/screenshots/*.py are scenarios: a setup() that builds a tree (and optionally after(context)). scripts/screenshot.py opens Blender on it and saves the window; scripts/docs_screenshots.py regenerates all docs images.