Warnings in tests

[opotowsky]

I have two types of warnings I see here:

(1)

and (2)

I don't want to fix (1) without some advice because that's an issue with the code itself (in multiple locations, but it's the same SORT of problem and will likely be handled in the same way in all locations).

And (2), I don't know HOW to fix.

Help?

[opotowsky]

These are the remaining ~23 warnings in the ARMI unit tests so we are fairly close to being warning free 💃

[john-science]

Well, the source of number (2) above is the legend:

armi/armi/utils/reportPlotting.py

Line 247 in 66551d5

plt.legend()

If you remove that line, those warnings go away.

When I make the plot from the test test_xsHistoryVsTime(), the legend has missing symbols:

Okay, the problem is the xsType variable that we inject into the legend, that comes from crossSectionGroupManager.getXSTypeLabelFromNumber(typeNum):

armi/armi/utils/reportPlotting.py

Lines 239 to 241 in 66551d5

	xsType = crossSectionGroupManager.getXSTypeLabelFromNumber(typeNum)
	color = next(colors)
	plt.plot(times, burnups, color + ".", label="Type {0} XS".format(xsType))

For me, in that test, the xsType is an empty string. Which is... wrong? But I'm not sure why that shows up as anything funny in the legend.

[john-science]

Okay, the problem for your number (2) here is our tests are stupid. They are providing invalid type numbers:

armi/armi/utils/tests/test_reportPlotting.py

Lines 121 to 122 in 66551d5

	1: [[0, 1], [0, 2], [0, 3]],
	2: [[0, 5], [0, 6], [0, 7]],

That should read something more like:

        65: [[0, 1], [0, 2], [0, 3]],
        6570: [[0, 5], [0, 6], [0, 7]],

Or whatever. Just valid type numbers.

[ntouran]

Sending in a PR for the reportPlotting issue.

[john-science]

Update: Nick's PR fixed problem (2).

So, only (1) is left.

But if I were a gambling man, I would disagree with Arrielle here. I bet all these "divide by zero" warnings are not related. I bet there are many tests with silly default values chunked in, and those cause the warnings.

My guess is that each of these tests will have to be tweaked individually, so they have better defaults. Or, in some cases, a more complete reactor will have to be provided to the test.

[john-science]

Also, as a matter of "what to attack first", I want to focus on the warnings from the unit tests in Python 3.13 first.

example

[opotowsky]

But if I were a gambling man, I would disagree with Arrielle here. I bet all these "divide by zero" warnings are not related. I bet there are many tests with silly default values chunked in, and those cause the warnings.

My guess is that each of these tests will have to be tweaked individually, so they have better defaults. Or, in some cases, a more complete reactor will have to be provided to the test.

There are just a handful (not even) of places in the code where this needs to be handled, not the tests themselves. So it's not all the same root cause of warning, but I think we would resolve most of them in the same way (most are divide by 0 warnings)

[opotowsky]

hey it's down to 18 warnings tho! better place than I thought we'd be!