We are making some API changes that require us to regenerate our testing snapshots (we're using pytest-snapshot, another more familiar tool is snapshottest). Before pants the way that we did this is to pass the

--update-snapshot

option to the pytest CLI. Now that we're on pants I'm unsure which mechanisms, targets, goals we would use instead. I can run

./pants test package/tests/test_example.py -- --update-snapshot

but the new snapshots I want to save are inside the sandbox. So how would I use my

python_tests

targets to generate new

resources

targets (which are ultimately Python files which contain the snapshots)?

There’s a pytest args option for the python tests subsystem

We're able to pass in args to pytest without a problem. In our case this is a command line arg that we want to pass only when we're regenerating our screenshot files and not every time we run the test goal. When we do pass the command line arg it works as expected but inside the sandbox, the new files are inside the sandbox which isn't helpful because we want to persist our new snapshots as

resource

targets for the tests to use.

Ah. Sounds like you’re approaching plug-in territory to ship the generated files to dist/ where other exports go

Yeah, there is a ticket for this feature - https://github.com/pantsbuild/pants/issues/11622

thanks a lot @happy-kitchen-89482 , it was quite easy to get it working for Helm. Unfortunately I’m not that familiar with this feature in pytest and how users usually interact with it but happy to give a hand if someone wants to have a go at it

Yeah, I will take a look at Alonso's code on the helm snapshots. I'm not using it on a pants repo, but there's another similar use-case on the python side for test cassettes as well. Basically I'd like to be able to find new/changed artifacts that are created alongside the

python_test

targets and move them out of the sandbox and into the repo so they can become

resource

targets. @witty-family-13337 it sounds like the helm snapshots are doing something similar? Depending on the testing plugin / config you're using these files could show up in folders with different naming conventions and could also be YAML, Python, or other types of files. I'm not familiar with helm very much but it sounds like snapshots are a supported tool by helm itself and not implemented as a third party plugin. This seems more convenient as the file naming pattern / format would be consistent.

yeah, you’re more or less spot on. The Helm snapshots are a feature of a Helm plugin named

unittest

. In this plugin’s context, snapshots are YAML files that live in a folder with named

__snapshot__

(no option to change it). The implementation in Pants rests in 3 pilars: 1. Tailor: Being capable of detecting the existence of snapshot folders and generate

resource

targets for them. 2. Dependency Inference: On the presence of

resources

targets in a snapshot folder (relative to the test sources), then establish a dependency relationship between the test sources and the snapshot resources (the user should still be capable to explicitly negate that dependency like in other cases supported by Pants) 3. Generate Snapshots: This is a new goal that would trigger the generation of new or updates test snapshots. If new snapshots are generated, then a new tailor run may be needed to get the

resources

targets generated again.

Great, super helpful. Thank you!

You can start with just #3, and require manual addition of resources targets and the deps on them, as a first step.