Any recommended ways to update the executor of a pre-defined or imported Electron object?

[Andrew-S-Rosen]

Is there a convenience function in Covalent to update the executor of a pre-defined Electron? If there's not, would it be worth it to make one?

I know that, for an electron named myelectron, you can do something like myelectron.electron_object.metadata["executor"] = "local", but I wasn't sure if that sets all the necessary variables (and if it's recommended).

The reason I'm asking is that I'm developing a library of pre-defined workflows that users will import into their own codes. I'd like to ensure that if a user imports an Electron object from my package, they can set the executor to whatever they like. The alternative is for me to define functions and have the user wrap each function in a ct.electron() "manually" so they can set the executor when they define the Electron, but that's slightly less preferred for my use case.

This is somewhat related to #1586 and #1587, but I've modified the scope of my question a bit.

[santoshkumarradha]

hey @arosen93 , this is a long backlogged item on our plate. Ideally the best UX is to have something like

UX1

from my_package import energy_electron # imports a direct electron object

energy_electron.executor=MyExecutor()

Could @kessler-frost confirm if using energy_electron.metadata["executor"]=MyExecutor() would be a valid solution? I'm uncertain if it would properly set everything needed. I recall encountering some issues with UX1 in the past, but perhaps they have since been resolved and we can revisit the matter.

Meanwhile I am just curious if using a UX2 would be a good fit for creating custom electron packages.

UX2

from my_package import energy_electron
small_cell=energy_electron(**parameters,executor=MyExecutor(num_cpu=2)) # Returns an electron function 
super_cell=energy_electron(**parameters,executor=MyExecutor(num_cpu=45)) # Returns an electron function 

@ct.lattice
def workflow():
    energy_supercell=super_cell("NaCl") #different resource electron
    energy_smallcell=small_cell("NaCl") #different resource electron

In certain scenarios, UX2 may be more suitable as it allows for the creation of multiple electron instances with unique parameters through a factory pattern. This is not achievable in UX1, which is designed to return a singleton electron instance with a specific tied instance. Infact our initial interest in UX1 was coming in from a perspective of changing executor inside an interactive notebook after electrons have been defined.

LMK your thoughts.

[kessler-frost]

I think we can make both of the UX's work with only minor changes.

[Andrew-S-Rosen]

You're really making my day now with all this uplifting news, @kessler-frost!

Issue opened here: #1647

[kessler-frost]

How about a UX2.5 instead of UX2:

from my_package import energy_electron
small_cell=energy_electron(executor=MyExecutor(num_cpu=2))(**parameters) # Returns an electron function 
super_cell=energy_electron(executor=MyExecutor(num_cpu=45))(**parameters) # Returns an electron function 

@ct.lattice
def workflow():
    energy_supercell=super_cell("NaCl") #different resource electron
    energy_smallcell=small_cell("NaCl") #different resource electron

This way,

• It will not remove "executor" (or any other field if we decide to add more) from a possible set of argument names
• Will probably be easier to implement

[kessler-frost]

Or maybe even swap the positions:

small_cell=energy_electron(**parameters)(executor=MyExecutor(num_cpu=2)) # Returns an electron function 
super_cell=energy_electron(**parameters)(executor=MyExecutor(num_cpu=45)) # Returns an electron function 

[Andrew-S-Rosen]

I think pretty much any compact solution is probably a good one, so long as it's documented! :D 2.5 (or 2.6 lol) both look good to me.