TST: Relax test_aa_high_precision#11564
Conversation
| assert_allclose(moon_aa.az, TARGET_AZ, atol=0.1*u.uas, rtol=0) | ||
| assert_allclose(moon_aa.alt, TARGET_EL, atol=0.1*u.uas, rtol=0) | ||
| assert_allclose(moon_aa.distance, TARGET_DISTANCE, atol=0.1*u.mm, rtol=0) | ||
| assert_allclose(moon_aa.az, TARGET_AZ, atol=0.1*u.uas, rtol=3.7e-10) |
There was a problem hiding this comment.
@pllim - could you increase the absolute tolerance by a factor 10 instead? That is more in the spirit of the test.... Sorry for the bother!
There was a problem hiding this comment.
Factor of 10 is 1 * u.uas, right? That won't cut it.
>>> moon_aa.az
<Longitude 15.03267352 deg>
>>> TARGET_AZ
<Quantity 15.03267351 deg>
>>> moon_aa.az - TARGET_AZ
<Angle 5.44987699e-09 deg>
>>> (moon_aa.az - TARGET_AZ).to(u.uas)
<Angle 19.61955718 uas>There was a problem hiding this comment.
I updated the tests to relax atol to the smallest values that would make the comparison pass (at least locally).
There was a problem hiding this comment.
OK, that sounds good - we definitely want the tests to pass again!
There was a problem hiding this comment.
They did pass, but I have no idea what this change means. Is the moon not where it is supposed to be?
There was a problem hiding this comment.
It is just that with the slightly more accurate way erfa deals with polar motion, implied coordinates differ a bit.
pllim
force-pushed
the
shhhh-round-earth
branch
from
57a2330 to
721273b
Compare
|
Let's get this in! |
|
This test did not exist in v4.0.x so changing milestone |
3 participants
Description
This pull request is to fix #11527 unless someone has a better solution.
If this does not need backporting, feel free to re-milestone.
Note to reviewers: Check the "allowed failure" job.