`flobject.get_root()` is returning ~2 MB of data through gRPC

[raph-luc]

Call: GRPC_TRACE: rpc = /ansys.api.fluent.v0.settings.Settings/GetStaticInfo, request = {'root': 'fluent'}

MWE:
Execute this with logging enabled

import ansys.fluent.core as pyfluent
from ansys.fluent.core import examples

import_filename = examples.download_file("ablation.msh.h5", "pyfluent/examples/Ablation-tutorial")

session = pyfluent.launch_fluent()

session.tui.file.read_case(import_filename)

session.field_data.get_scalar_field_data.field_name.allowed_values()

session.solution.initialization.hybrid_initialize()

then check logs

Edit: this does not only happen when there is an API mismatch, this happens even with a built package/wheel installation with full codegen files

Tested using latest Fluent v241 container (2023-07-17T07:44:48Z)

Edit 2: Seems to be triggered by

pyfluent/src/ansys/fluent/core/session.py

Lines 161 to 163 in daf072d

	self.field_data = FieldData(
	self._field_data_service, self.field_info, _IsDataValid(self.scheme_eval)
	)

which calls ansys.fluent.core.services.scheme_eval.SchemeEval.scheme_eval

pyfluent/src/ansys/fluent/core/services/scheme_eval.py

Lines 320 to 326 in daf072d

	S = Symbol # noqa N806
	val = (
	S("eval"),
	(S("with-input-from-string"), input, S("read")),
	S("user-initial-environment"),
	)
	return self.eval(val)

due to the logs of the grpc request

GRPC_TRACE: rpc = /grpcRemoting.SchemeEval/SchemeEval, request = {'input': {'list': {'item': [{'sym': 'eval'}, {'list': {'item': [{'sym': 'with-input-from-string'}, {'str': '(data-valid?)'}, {'sym': 'read'}]}}, {'sym': 'user-initial-environment'}]}}}

which receives the response

GRPC_TRACE: response = {'output': {'b': False}}

[seanpearsonuk]

@raph-luc I see that you merged a PR. Is this issue closable?

[raph-luc]

@seanpearsonuk The "bigger picture" issue is that we are requesting the full flobject static info often (probably more often than we need to, would require researching whether we can optimize/reduce the number of requests).

In my understanding, as more objects are added to the settings API, the larger these full static info calls will become. Even though the gRPC messages are compressed (2 MB was the uncompressed value, considering compression it was around 1 MB if I recall correctly), this could/will become a performance issue at some point, particularly if Fluent is running remotely from PyFluent.

The issue that actually prompted me to look into this, was that these very large messages were causing logging issues, and that has been resolved by the mentioned PR.

[mkundu1]

@raph-luc @seanpearsonuk Currently, the settings API static info is requested even though we have the latest generated files from codegen, because we are comparing the hashes to check for a potential mismatch in the API. We can make the hash comparison a dev feature, i.e. we skip the hash comparison if the connected Fluent is from a released version and there are generated files from codegen for that Fluent version.

Regarding the size of the static info response, we can change it to a streaming response if needed, let's discuss that on Fluent server side.

[seanpearsonuk]

@mkundu1 With the current state of play, do you envisage that improvements can be made in this area?

[mkundu1]

@seanpearsonuk #3399 has removed the static-info query during runtime, closing.