remove Hamilton liquid classes

[rickwierenga]

Liquid classes in the Hamiltonian sense are confusing, especially in PLR: https://labautomation.io/t/how-does-pylabrobot-handle-liquid-classes/977

A lot of code in STAR looks like this:

flow_rates = [
      op.flow_rate or (hlc.aspiration_flow_rate if hlc is not None else 100)
        for op, hlc in zip(ops, hamilton_liquid_classes)]

Hamiltonian liquid classes are effectively dicts of parameters.

We should do the following instead:

lh.aspirate(..., **hamilton_liquid_class.generate_kwargs())`

>>> hamilton_liquid_class.generate_kwargs()
{
  "flow_rate": 42,
  "clot_detection_height": 69,
  ...
}

so that we can remove all references to Hamilton liquid classes in STAR/Vantage backends, thus simplifying them a lot. HamiltonLiquidClass can simply be a TypedDict.

(I vaguely recall discussing this before, possibly in a private message, can't find a public issue/forum post now)

[rickwierenga]

one issue with making it a TypedDict is that volume correction becomes slightly more awkward, and the user needs to explicitly say

lh.aspirate(vols=[hlc.correct(original_volume)])

but i think for sensitive liquid handling you'll want to fine-tine the volumes anyway, and this method is a lot more transparent than doing it somewhere deep in STAR where it's hidden

[rickwierenga]

first PR: #248

[BioCam]

one issue with making it a TypedDict is that volume correction becomes slightly more awkward, and the user needs to explicitly say

lh.aspirate(vols=[hlc.correct(original_volume)])

I agree; this puts enormous technical debt on the end user.

This way every user must constantly think about the actual piston_volume, i.e. the "corrected_volume", one has to move to achieve the target_liquid_volume.

The excellent idea behind standard liquid classes is to not have worry about this.
It makes reading and writing complex protocols much harder.

Ideally we want a system/advanced_liquid_class that is more than a dictionary, i.e. an object that corrects not just for volume but corrects all aspirate/dispense attributes... because why would the same blow_out_volume for 1 ul of 50% glycerol work the same way for 1000 of 50% glycerol.

This current alternative doesn't seem to address this.

[rickwierenga]

but with liquid classes, you would still have to worry about this, because the provided liquid classes aren't good enough.

yes, we don't have an 'advanced liquid class' yet.