PlantArchitecture: adjusting maize shoot parameters seems not to work

[jandleuven]

I am using pyhelios version: 0.1.28
I am trying to change the maize flowering time by adjusting the exposed Python parameters phyllochron_min and max_nodes on the mainstem shoot type:

plant = PlantArchitecture(context)
plant.loadPlantModelFromLibrary("maize")

params = plant.getCurrentShootParameters("mainstem")
params["phyllochron_min"] = {"distribution": "constant", "parameters": [2.0]}
params["max_nodes"] = {"distribution": "constant", "parameters": [17]}
plant.defineShootType("mainstem", params)

But when I do that (and still keep the maize default values 2.0 and 17.0), tassels appear on every node (and ears are not formed). Most likely the definition of the MaizePhytomerCreationFunction in C++ gets lost when using the statement plant.defineShootType("mainstem", params). That would mean you cannot change any of teh maize parameters from Python. Is that correct? Any workaround?

[aneeshpanoli]

shoot_types["mainstem"] gets overwritten at the end of the round trip — that's where it breaks.

When shoot_types["mainstem"] serializes on the C side it drops both phytomer function pointers. Then the object crosses over to the Python side, gets modified, and makes the trip back to the C side and overwrites shoot_types["mainstem"], leaving 2 null pointers for the functions (phytomer_creation_function|MaizePhytomerCreationFunction, and the callback).

The solution would be to capture the function pointers from shoot_types["mainstem"] just before overwriting and then put them back. This would preserve them without exposing them.

Happy to pick this up if it helps.

[jandleuven]

@aneeshpanoli
Thanks very much for looking into this. Is this something I can avoid by modifying my Python script? Or does it need changes in pyHelios (in Python wrappers or in C++)?

[bnbailey-psl]

Thanks for the clear report. You were right that the phytomer creation function was being lost: it's a raw C function pointer that can't survive the JSON round-trip, so defineShootType was rebuilding 'mainstem' with a null hook and dropping the logic that places ears at nodes 9–11. It has been fixed.

Unfortunately it did require a C++ fix without a Python work-around. But I have a number of changes bending so should be good to push out a new version soon.

[aneeshpanoli]

@jandleuven Unfortunately none. The destructive call is defineShootType(...), and that is the only shoot parameter door from Python to C++. So, the fix has to land in the C++ wrapper.