refactor(tests): simplify test code

[lxl66566]

changes:

• merge validity_test and stdlib_test logic into an inner function script_test: the test logic are same, it's better not to use same code
• rewrite extract_output to more convenient form (with test)

[Mte90]

I see that #362 include some stuff that are in this PR.
So what is the right one to review? maybe the other one can be closed?

[b1ek]

I see that #362 include some stuff that are in this PR. So what is the right one to review? maybe the other one can be closed?

this PR depends on the other one, means the author has included that PR's commits in this one. just ignore it for now and wait for 362 to be merged

[Mte90]

Or we can merge this one as include more stuff.

[lxl66566]

i rebased this pr, it's not related to another pr now, that may be more convenient

[Mte90]

@lxl66566 can you fix the conflicts?

[lxl66566]

@lxl66566 can you fix the conflicts?

rebased

[Ph0enixKM]

I'd rather not. When changing some internal functionality - I want to see all the validity checks green before I check the stdlib.

• The standard library tests are testing the functioanlity of the standard library functions.
• The validity tests check whether the compiler works properly.

Why keep the separate?

1. To separate the library noise from the compiler itself - I want to check if what I changed messed up with the core language itself and I don't want to worry about some library functions fail because I'm currently on macOS and some command relies on Linux version of the command API.
2. To have a way to run just the compiler checks - sometimes library tests can fail and that's okay (some commands might be not installed on the system). Compiler tests cannot fail at all.

Do you agree with me?

[Ph0enixKM]

I think that we shouldn't move validity (the compiler validity) tests to the stdlib (standard library) tests.

[lxl66566]

I think that we shouldn't move validity (the compiler validity) tests to the stdlib (standard library) tests.

I got you. I think we can put the test logic in script_test(input) in mod.rs, and just use stdlib_test(input), validity_test(input) as two wrappers to call inner script_test(input). That can avoid simply repeating the same code.

And then, put stdlib_test(input) and validity_test(input) in two files, so you can pick any one you like to test.

How do you think of that?

[KrosFire]

Taking this opportunity. Why don't we start using insta? We wouldn't have create output txt files. We would just approve generated ones

[Mte90]

Taking this opportunity. Why don't we start using insta? We wouldn't have create output txt files. We would just approve generated ones

Right now with the latest changes there are just a couples of .txt files as we removed the majority.

[Mte90]

Can you align with the latest changes?
So @Ph0enixKM can review it :-)

[lxl66566]

master now in #300 didn't pass the test.

[Mte90]

A pr fixed that issue.

[Ph0enixKM]

Bug detected

[Ph0enixKM]

All good now