Dim sort

[MichaelTiemannOSC]

Adapt PR#1841 to the new Pint formatter.

• Closes Override units display order (1 kW·h displayed as 1 h·kW) #1841
• Executed pre-commit run --all-files with no errors
• The change is fully covered by automated unit tests
• Documented in docs/ as appropriate
• Added an entry to the CHANGES file

TODO: show we can override default in registry init file.

[codspeed-hq]

CodSpeed Performance Report

Merging #1926 will not alter performance

_{Comparing dim-sort (32f543d) with master (3cc2d36)}

Summary

✅ 439 untouched benchmarks

[MichaelTiemannOSC]

This test creates a UnitsContainer with meter inserted before kilogram:

    def test_unit_formatting(self, subtests):
        x = self.U_(UnitsContainer(meter=2, kilogram=1, second=-1))
        for spec, result in (
            ("{}", str(x)),
            ("{!s}", str(x)),
            ("{!r}", repr(x)),
	    (
                "{:L}",
                r"\frac{\mathrm{kilogram} \cdot \mathrm{meter}^{2}}{\mathrm{second}}",
            ),
            ("{:P}", "kilogram·meter²/second"),
            ("{:H}", "kilogram meter<sup>2</sup>/second"),
            ("{:C}", "kilogram*meter**2/second"),
            ("{:Lx}", r"\si[]{\kilo\gram\meter\squared\per\second}"),
            ("{:~}", "kg * m ** 2 / s"),
            ("{:L~}", r"\frac{\mathrm{kg} \cdot \mathrm{m}^{2}}{\mathrm{s}}"),
            ("{:P~}", "kg·m²/s"),
	    ("{:H~}", "kg m<sup>2</sup>/s"),
            ("{:C~}", "kg*m**2/s"),
        ):
            with subtests.test(spec):
                assert spec.format(x) == result

The UnitsContainer formatter, by default, sorts these alphabetically, putting kilogram before meter**2. But if the dimensional sort function is None, it imposes no sort, leaving insertion order to rule. Which leaves meter**2 before kilogram.

This argues for the idea that we must default sorting compound units with sorted (by alpha).

[MichaelTiemannOSC]

I don't know why that one test failed--looks like a CI hiccup.

[hgrecco]

I don't know why that one test failed--looks like a CI hiccup.

I can fix that in my computer. But how do you think we should merge #1864 and #1926 (this?) Do want to merge them into one and I merge them?

[MichaelTiemannOSC]

I think this is the one PR to merge (the other is obsoleted by this one). We can certainly add some more tests after merging that fleshes out how complex multi-unit dimensions should be sorted, but that will only change the inner operation of the dim_sort function, not the interfaces (I think).