add ventilation coefficient (Pruppacher & Rasmussen 1979 & Froessling 1938) + Reynolds number attr + air dynamic viscosity formulae + unit tests (incl. drop size changes upon ventilation-in-parcel-below-cloud vs. no-ventilation) + refactor of env vars __getitem__ logic

[kaitlyn-loftus]

First attempt to add ventilation coefficient from Pruppacher & Rasmussen (1979)).

Added methods for air dynamic viscosity, particle Reynolds number, and air Schmidt number to support ventilation coefficient calculation.

Tried to add basic tests for dynamic viscosity and ventilation coefficient calculation but don't think I activated them...

[slayoo]

@kaitlyn-loftus, thanks!!!

For all PRs, we run the CI workflow in several steps, on of the first featuring pylint check (https://pylint.readthedocs.io/). Here it complains a bit:

************* Module PySDM.physics.ventilation.pruppacher_rasmussen_1979
PySDM/physics/ventilation/pruppacher_rasmussen_1979.py:2:0: C0301: Line too long (110/100) (line-too-long)
PySDM/physics/ventilation/pruppacher_rasmussen_1979.py:3:0: C0301: Line too long (107/100) (line-too-long)
PySDM/physics/ventilation/pruppacher_rasmussen_1979.py:5:0: C0301: Line too long (182/100) (line-too-long)
PySDM/physics/ventilation/pruppacher_rasmussen_1979.py:69:0: C0301: Line too long (131/100) (line-too-long)
PySDM/physics/ventilation/pruppacher_rasmussen_1979.py:83:0: C0301: Line too long (122/100) (line-too-long)
PySDM/physics/ventilation/pruppacher_rasmussen_1979.py:17:4: C2401: Function name "calcηairEarth" contains a non-ASCII character, consider renaming it. (non-ascii-name)
PySDM/physics/ventilation/pruppacher_rasmussen_1979.py:37:42: C2401: Argument name "ηair" contains a non-ASCII character, consider renaming it. (non-ascii-name)
PySDM/physics/ventilation/pruppacher_rasmussen_1979.py:37:49: C2401: Argument name "ρair" contains a non-ASCII character, consider renaming it. (non-ascii-name)
PySDM/physics/ventilation/pruppacher_rasmussen_1979.py:53:15: C2401: Argument name "ηair" contains a non-ASCII character, consider renaming it. (non-ascii-name)
PySDM/physics/ventilation/pruppacher_rasmussen_1979.py:53:25: C2401: Argument name "ρair" contains a non-ASCII character, consider renaming it. (non-ascii-name)
PySDM/physics/ventilation/pruppacher_rasmussen_1979.py:121:8: R1705: Unnecessary "else" after "return", remove the "else" and de-indent the code inside it (no-else-return)
************* Module tests.smoke_tests.no_env.pruppacher_rasmussen_1979.test_dynvisvT
tests/smoke_tests/no_env/pruppacher_rasmussen_1979/test_dynvisvT.py:28:0: C2401: Function name "test_ηvT_figA5" contains a non-ASCII character, consider renaming it. (non-ascii-name)
tests/smoke_tests/no_env/pruppacher_rasmussen_1979/test_dynvisvT.py:28:23: C2401: Argument name "η" contains a non-ASCII character, consider renaming it. (non-ascii-name)
tests/smoke_tests/no_env/pruppacher_rasmussen_1979/test_dynvisvT.py:30:4: C2401: Variable name "η_test" contains a non-ASCII character, consider renaming it. (non-ascii-name)

For this reason, most of the tests were not run.
To check how to execute pylint locally, see: https://github.com/open-atmos/python-dev-hints/wiki/Pylint-FAQ

The smoke_tests folder is meant for tests that cannot be called "unit tests" for they involve executing multiple units of code. Here, I'd say you've implemented exactly what unit test should look like, so I'd suggest moving the new tests into PySDM/tests/unit_tests/physics. Any '.py file the begins with test_ will be included in the test run, there is no need to activate it.

To run the unit tests locally, executing pytest tests/unit_tests should be enough.

Thanks again!

[slayoo]

Thank you!
The new test was executed and passed! (part of thenojit_and_codecov job which runs with Numba JIT turned off):
https://github.com/open-atmos/PySDM/actions/runs/7966038711/job/21746648231?pr=1282#step:4:808

[kaitlyn-loftus]

Great!! Thanks for explaining. I originally ran the pylint tests before staging the new files, so none of them got checked 😅

[slayoo]

Thinking out loud, a next step here could be to add a new super-particle attribute: Reynolds number.
The calculation of it would go to backends which would call the newly introduced formula.
Then, within condensation/evaporation dynamic, this attribute would be available in case Formulae was instantiated with a non-neglect ventilation specified... please give me a few days to come up with a draft code - will push it into this PR's branch when ready.

[slayoo]

@kaitlyn-loftus, in the above commit 9c2269f, there are some refactors towards making the ventilation coefficient being used inside condensation/evaporation solver. Added also more unit tests and a simpler "classic" formula from Froessling 1938 for reference. Please have a look if the changes look OK? Next step on my side: use the ventilation coefficient in the condensation solver - stay tuned, thanks

[kaitlyn-loftus]

@slayoo these changes look good. Thanks for fixing the air density to include the water vapor contribution; I had meant to flag that as a todo. The only potential issue I see is how the ventilation_coefficient method in PySDM/physics/ventilation/pruppacher_rasmussen_1979.py would handle sqrt_re_times_cbrt_sc=const.PRUPPACHER_RASMUSSEN_1979_XTHRES. This looks to me like it would give NaN (unless there is some clever handling elsewhere in the code).

[slayoo]

The only potential issue I see is how the ventilation_coefficient method in PySDM/physics/ventilation/pruppacher_rasmussen_1979.py would handle sqrt_re_times_cbrt_sc=const.PRUPPACHER_RASMUSSEN_1979_XTHRES. This looks to me like it would give NaN (unless there is some clever handling elsewhere in the code).

Very valid point! Thanks, @kaitlyn-loftus
Here's a fix with a new test added checking behavior at the threshold value: 4917da4

[kaitlyn-loftus]

I think this will still give 0 at the threshold, instead of ~1.02. Is np.heaviside friendly with the CPU and GPU accelerators?

[slayoo]

I think this will still give 0 at the threshold, instead of ~1.02. Is np.heaviside friendly with the CPU and GPU accelerators?

Thanks, @kaitlyn-loftus !
Hopefully fixed in f886fec

We don't have yet any tests covering GPU for this, so let's address GPU compatibility when we will get there...
Added an assert for the value (which seems to be around 1.2)

[slayoo]

@kaitlyn-loftus just a quick update: here's a fresh PR introducing a refactor of the CPU condensation logic which is meant to make the introduction of the ventilation factor there much simpler - #1291

[slayoo]

@kaitlyn-loftus, the a40ca41 commit above contains draft of code for actually using the ventilation coeffs in condensation integrator. This includes a basic test depicting that evaporation of a single droplet in subsaturated env proceeds faster if ventilation if taken into account: https://github.com/open-atmos/PySDM/blob/a40ca41377ecb8e57f68d83cf81e605071511147/tests/unit_tests/dynamics/condensation/test_ventilation.py

Stay tuned for updated GPU code.

[slayoo]

@kaitlyn-loftus, CI green! :)
Please have a look, happy to merge soon and start working on 1D and 2D examples featuring ventilation!
(also, this will be super important for the isotope fractionation developments)

[kaitlyn-loftus]

great news!

[slayoo]

OK, I'm merging then, and will follow right away with a new release.
Thank you @kaitlyn-loftus!!!