[GiD model] Submodel part naming conflict

[mpentek]

When working with multiple geometry layers and groups for better handling of complex geometry, I have the following situation:

So I will end up with submodel parts when using Kratos with names such as:
FluidModelPart.Drag_mq/1
which is internally used for output (file) generation. I find the choice of the "/" as subgroup/child separator not the best choice. Was this intended to be like this, does it have to be like this? Is it something that comes from the Kratos problemtype? @rubenzorrilla I assume that a lot of processes might encounter errors due to special character such as "/" included in names.

[rubenzorrilla]

Honestly, I have never used subgroups so far... In any case I agree that the slash is not the best character.

@jginternational can we control this?

[jginternational]

From the 'kratos point of view' there are no groups, we'll have submodelparts

You are right -> The name of the submodelparts whose group is child is written like this:
"FluidModelPart.Outlet2D_Outlet/Child"
Maybe we can change it to this if we don't like the / as separator:
"FluidModelPart.Outlet2D_Outlet-Child"

Spaces are changed for underscores _
Child slashes could be replaced for regular score -

[philbucher]

FluidModelPart.Outlet2D_Outlet-Child sounds good to me 👍 , it should work

[rubenzorrilla]

The - is fine for me as well.

However, I think that the subgroups in GiD (1, 2, 3, 4 and 5 in @mpentek snapshot) should be subsubmodelparts in the *.mdpa. Is this possible @jginternational ?

[jginternational]

Nope @rubenzorrilla
A submodelpart is created in the join of a Condition (on the problemtype) with a group
example:
If I assign a group to an Inlet I'll have the submodelpart Inlet_group
If I assign a group to an Inlet twice (same group, different time intervals) I'll have only 1 submodelpart Inlet_group

If I assign a subgroup group//subgroup to an inlet, the submodelpart created is expected to be
Inlet_group-subgroup

• I hate the polysemy of the word Condition

[mpentek]

For me the main concern for now is the format, meaning avoiding separators which are generally used for folder path and so on. If this were to be changed from "/" to "-", would it cause problems with compatibility, tests and so on @jginternational? I think it would by anyway worth it in the long run.

The discussion of model/submodel/subsubmodel is perhaps another aspect. Here @rubenzorrilla and @philbucher can say more.

[jginternational]

@mpentek I'm merging that branch soon if there are no problems!

[mpentek]

Re-opening this issue, as with the most current setup (Win10 + GiD15.0.1 + Kratos GiD problemtype - from here) I have the same ongoing issue: the naming will be with "/" or "//".

Am I missing something? I had the impression this was changed.

[philbucher]

@jginternational can you please comment if this is solved?
This breaks many outputs in Kratos!

[jginternational]

sorry. The requested changes have been pushed to master after checking all the tests.

I though I wrote this yesterday, but it seem that I forgot to click on the green button :(

[philbucher]

coolt thx!

[sunethwarna]

Hi, Is there a way to avoid using "-" when naming model parts, submodel parts when used with groups? Because hdf5 cannot write names with "-" (no idea why tho). @jginternational . These "-" are automatically generated when you have groups within groups :/

[mpentek]

"-" was the proposal to get away from "//", which was conflicting with path separators.