Fix complex parsing signs

[ComplexSymbol]

Changes

I tried to fix a parsing error in the py/parsenum.c file.

Tests:
Before: print(complex('-9e-17+1j')) # prints (-9e-17-1j) wrong
After: print(complex('-9e-17+1j')) # prints (-9e-17+1j) correct

It's pretty simple, I just made it so it resets dec_neg upon restarting the parse for complex numbers so it doesn't carry through to the imaginary part. dec_neg gets applied to the real part before moving on to the imaginary, so resetting it doesn't change the sign of the real part.

Other stuff

Note: The contributions listed here on my fork are at the main branch, the other (doubles-cmath) branch on my fork is for my project, not contributing.

This is my first time contributing to anything! I'm not sure whether to submit this to main or the latest branch.

[dhalbert]

Thanks for this! Submitting on main is fine: we are not planning further 9.2.x releases except for egregious errors.

This code comes from MicroPython, so it would be appropriate to submit this upstream, to https://github.com/micropython/micropython. It's worth looking to see if the code there has already been fixed or is different in some way.

[Neradoc]

Yeah I was looking at MP and I see the same in the code, and same results when trying on 1.25.

>>> complex("1+j")
(1+1j)
>>> complex("-1+j")
(-1-1j)

[ComplexSymbol]

Thanks for this! Submitting on main is fine: we are not planning further 9.2.x releases except for egregious errors.

This code comes from MicroPython, so it would be appropriate to submit this upstream, to https://github.com/micropython/micropython. It's worth looking to see if the code there has already been fixed or is different in some way.

Should I submit a PR over there too then?

[dhalbert]

Should I submit a PR over there too then?

Yes, submit there first. If they accept it (and they might change it a bit), then we'll take their change as it is merged there, so the sources match. If they don't do it promptly, we can merge it here and then catch up with what they do.

[ComplexSymbol]

Should I submit a PR over there too then?

Yes, submit there first. If they accept it (and they might change it a bit), then we'll take their change as it is merged there, so the sources match. If they don't do it promptly, we can merge it here and then catch up with what they do.

Ok, thanks!

[dhalbert]

Another thing to do is to add a test: add it to tests/float/complex1.py. You can just add a few more cases to the set of print statements at the beginning. The standard test automation runs the test in micropython and in regular CPython, and the results should be the same. (It's also possible to specify the expected results with a .exp file, but I think the first case should be fine here for simple floats with e notation that can be represented exactly.)

[ComplexSymbol]

Another thing to do is to add a test: add it to tests/float/complex1.py. You can just add a few more cases to the set of print statements at the beginning. The standard test automation runs the test in micropython and in regular CPython, and the results should be the same. (It's also possible to specify the expected results with a .exp file, but I think the first case should be fine here for simple floats with e notation that can be represented exactly.)

There are already tests there that should've caught this mistake, not sure why they weren't noticed... There are missing tests for exponential notation and complex numbers, should I add those to to the complex1.py tests or somewhere else?

[dhalbert]

There is no test like complex("-1+1j") (negative followed by positive) in complex1.py. That seems to me to be just an oversight.

It would be fine to add more tests, and that's the right file to add them in. Won't hurt.

[ComplexSymbol]

There is no test like complex("-1+1j") (negative followed by positive) in complex1.py. That seems to me to be just an oversight.

It would be fine to add more tests, and that's the right file to add them in. Won't hurt.

Whoops, I thought I saw something like that, anyways I'll go ahead and add those. Thanks!