Allowing changing of timestep size after starting solving?

[TessellateDataScience]

I'm liking the idea of using Python to control my OpenFOAM case setup & solving (and keeping everything in Python for ease-of-use). However, I have an issue where I can't change the timestep size after an arbitrary time. Is there a way to do this or should I dive into modifying the code (if so where should I go)? Code below shows how I'm currently trying to do this:

from foamlib import FoamCase
case = FoamCase("./")

# run startup with relaxed values
case.control_dict["deltaT"] 		= 0.0002
case.control_dict["endTime"] 		= 0.001
case.control_dict["writeInterval"] 	= 0.001
case.run()

# run most of computation with intended values
case.control_dict["deltaT"] 		= 0.001
case.control_dict["endTime"] 		= 0.002
case.control_dict["writeInterval"]	= 0.002
case.run()

Note this using numerous cores (i.e. parallised via decomposePar).

[gerlero]

@TessellateDataScience your code snippet should work (does it not?). For what it's worth, I've done things just like this in the past. Just make sure that the controlDict is configured to always (re)start the simulation from the latest saved time (e.g. you can do case.control_dict["startFrom"] = "latestTime" at the beginning)

[TessellateDataScience]

Thanks Gabriel. OF doesn't pickup use the changed time-step value in controlDict due to time data specifying the old TS value (e.g. in ./processor0/0.001/uniform/time). Essentially I've implemented following as a workaround:

cmd0 = "sed -i -e 's/deltaT          0.0002/deltaT          0.001/g' ./processor*/"
cmd1 = "/uniform/time"
os.system(cmd0 + str(startupEndTime) + cmd1)

[gerlero]

OF doesn't pickup use the changed time-step value in controlDict due to time data specifying the old TS value (e.g. in ./processor0/0.001/uniform/time).

Oops, sorry! This part is new to me (as in, I've used this trick before, but never to change the simulation's deltaT).

I still wonder if there might be an easier way to accomplish this that doesn't involve editing all those files. In any case, let me know if you can think of a more elegant implementation and/or API for this, as it is a common enough use case that I would like to support directly in the package.