fix missing revnum attribute on pure-python Satrec

[plentydone]

Failure looks like this:

ERROR: test_export_tle_sgp4init (sgp4.tests.TestFunctions.test_export_tle_sgp4init)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/andrew/python-sgp4/.tox/py312/lib/python3.12/site-packages/sgp4/tests.py", line 252, in test_export_tle_sgp4init
    outline1, outline2 = export_tle(sat)
                         ^^^^^^^^^^^^^^^
  File "/home/andrew/python-sgp4/.tox/py312/lib/python3.12/site-packages/sgp4/exporter.py", line 89, in export_tle
    append(str(satrec.revnum).rjust(5))
               ^^^^^^^^^^^^^
AttributeError: 'Satrec' object has no attribute 'revnum'. Did you mean: 'elnum'?

Let me know if any adjustments are needed.

[brandon-rhodes]

Thanks for putting together this fix, complete with a test! It looks like a couple of tests are failing in CI, but I'll be happy to merge this once you get them fixed. Thanks!

[brandon-rhodes]

Could you move this down to just beneath self.classification? That's what it sits next to over in the C++ code, and would make it a bit easier for my eyes to find it later than up here with the dates.

[plentydone]

I think setting this to 0 here was a flawed approach, which I didn't see until I saw the failing CI tests. To make the patch less invasive, I adjusted my approach just to fix the error when exporting a TLE from a pure python Satrec initialized with sgp4init, as I'm not aware of anything else that needs it.

[brandon-rhodes]

I like how you didn't just make sure that export_tle() worked, but that the same satellite came back when re-parsed.

[plentydone]

Thanks! Updated to use assert_satellites_match which I hadn't noticed before.

[plentydone]

Sorry for the spam with the broken tests. I couldn't replicate locally for some reason, will figure out in the next day or two.

[plentydone]

Finally came back to this, saw what I did wrong in the tests, and fixed and expanded them. Hopefully it's all good this time!

[brandon-rhodes]

I have finally had time to circle back to this, and in case anyone else's code ever looks for .revnum, I think it makes the most sense for the module to simply guarantee that the attribute is always there:

6d89ae1

Also, I have written a test.sh test script that tests locally whether both the compiled and the pure-Python versions of SGP4 work. That way, we avoid needing to have tests with branches inside that do different things depending on whether the C++ successfully compiled.

[plentydone]

Looks great, thanks Brandon!

[brandon-rhodes]

I'm hoping to find time next week for a release, to get this feature out. Thanks again for pointing out the problem.