Command instance re-using old arguments on subsequent calls in test env

[yamalight]

I've encountered an odd behavior where subcommands in test env will re-use optional arguments from previous calls.
At the top level - you have a program and a subcommand with optional flags. Following calls behave oddly:

1. Call myapp subcommand --init (correctly passes init as argument to handler)
2. Call myapp subcommand --other 1234 (correctly passes other as argument to handler, but also includes init from first call)
3. Call myapp subscommand (passes previous values for init and other instead of empty args)

Is this behavior intended? If yes, how would one reset args between tests? The only way I've found so far was to monkey-patch subcommand _optionValues which feels wrong 😅

I've created a small repro case in stackblitz: https://stackblitz.com/edit/node-nsyb1h?file=app.test.js
Calling npm test executes it.
Uncommenting monkey-patching in the last test case makes it behave as expected (i.e. not pass any args when none given to parse function).

[shadowspawn]

Thanks for the detail and example.

The short answer is make new Commands for each test. i.e. both program and subcommand.

The code isn't written to allow multiple calls to parse, and there isn't a routine to reset state.

[shadowspawn]

See #1565 for some additional tips and link to an example.

[shadowspawn]

(Feel free to ask more questions, just pointing you in the right direction!)

[yamalight]

@shadowspawn gotcha! thanks for the clarification 👍
I'll close the issue since it's expected behavior.