FIX: correct multichannel couplings

[mmikhasenko]

Closes #87

A major problem has been found in implementation of Multichannel BW, in the way how couplings are computed.
The couplings themselves appear only for Flatte. So these values were modified to be consistent with the correct implementation.

• fix json
• fix julia
• fix python

🐛 Bug fixes

• Correct the serialized multichannel Breit-Wigner couplings and deserialize gsq as a coupling squared instead of an energy width.

✨ New features

• Render a Dalitz plot and the $m_{13}$ and $m_{23}$ mass projections (including per-resonance components) of the deserialized amplitude model.
• Verify the deserialized generic_function dynamics (Bugg Breit-Wigner) against the model definition.

⚙️ Enhancements

• Render individual dynamical expressions through unfold_definitions() so that their complete symbolic structure is visible.
• Assert that each deserialized distribution reproduces the checksums in the model definition, so that the documentation build fails on a mismatch instead of only reporting it in the validation table.
• Document the multichannel Breit-Wigner convention, with references to HadronicLineshapes.jl and ComPWA/ampform-dpd#199.

📝 Documentation

• Render figures as SVG with transparent backgrounds and neutral colours, so that they remain readable in dark mode.
• Remove the note block about the amplitude model not being deserialized correctly, which no longer applies.

🔨 Maintenance

• Require ampform-dpd>=0.4.1 and upgrade the lock files accordingly.

Squash commit messages

* DOC: improve rendering on dark mode
* DOC: remove note block about validation
* ENH: assert equivalence of checksums
* ENH: render sub-expressions with `unfold_definitions()`
* ENH: verify `generic_function`
* FEAT: render Dalitz plot and mass projections
* MAINT: upgrade to ComPWA/ampform-dpd@0.4.1

[redeboer]

MultichannelBreitWigner is imported, so since it was matched, the problem might be there as well

from ampform_dpd.dynamics import (
    BreitWigner,
    ChannelArguments,
    EnergyDependentWidth,
    MultichannelBreitWigner,
    P,
    SimpleBreitWigner,
)

These features are not released, see here. But good to be aware for future developments.

[mmikhasenko]

The problem should be somewhere here, then,

• definition of MBW
• function to create from dict

[mmikhasenko]

@redeboer could you please help understanding your definition,

m^2-s-imG
G = sum Gi
#
Gi = 1/m * g^2 * 2p/sqrt(s)

In the code, I do not recognize the Gi expression.

[mmikhasenko]

@redeboer do the tests at the PR check the validation points of json in python code?

I'm surprised to see that the json is modified, python code is not modified, but test button is green.

[redeboer]

There are some tests, it seems
ComPWA/ampform-dpd@0.2.0...0.2.1rc0#diff-37a837280345b38e5d2b0707adb33df844b8d44971fcae9fcb1aa340073d16cf
But could well be that they don't cover the part of the code that you're referring to.

[mmikhasenko]

JSON has cross check evaluation, it is surprising that these are not evaluated or pass, given that the lineshape for L1405 changed

[redeboer]

it is surprising that these are not evaluated

Not really. Like I said, this was coded together in an afternoon and is therefore only in a release candidate. Serialization is currently not our priority list.

[mmikhasenko]

This issue is looking forward for being resolved :)

[mmikhasenko]

ping @redeboer

Let's check if your MultichannelBreitWigner does the same as this,
https://github.com/mmikhasenko/HadronicLineshapes.jl/blob/main/src/shapes.jl#L22-L31

before merging, I would like to see,

1. json is changed, python code gives different results as before (as expected, or as julia)
2. the python code is fixed, dependencies are updated
3. python code gives the same result as expected, or as julia

[mmikhasenko]

Ok, Is it fixed now, @redeboer ?

[redeboer]

Has the serialization format been formalized?

[mmikhasenko]

I mean, I have seen some development, looks like you worked on it. So I am curious if it is resolved. Not related to the format.