MeshIO using new Properties and PropertyVector. #644
Conversation
| "MaterialIDs", MeshLib::MeshItemType::Cell, 1) | ||
| ); | ||
| if (!opt_material_ids) { | ||
| WARN("No material ids read."); |
There was a problem hiding this comment.
is this possible? and if so, is it still a valid mesh file?
There was a problem hiding this comment.
It should not be possible. But ...
|
Two general things:
this would require a lot of changes to the current read method and i really don't know if it is worth the effort for a legacy interface. i am just suggesting that the current implementation might crash in a number of cases. but it's also okay if we decide not to care about that .... |
|
Another thing: we currently don't have tests for this. I am not sure if we need some (legacy interface and all that). if you feel like it you could write some ... |
TomFischer
force-pushed
the
MeshIOUsingNewPropertiesImpl
branch
from
065a20d
to
9897935
Compare
|
In my opinion this is a legacy file format. I do not want to put much more effort than using the new data structures for properties, i.e. I do not want write test nor I want to improve the robustness. With the new commits I hope I addressed all your other comments. |
| elements.push_back(readElement(line_string, nodes)); | ||
| std::stringstream ss(line_string); | ||
| materials.push_back(readMaterialID(ss)); | ||
| elements.push_back(readElement(ss, nodes)); |
There was a problem hiding this comment.
no need to stop reading if readElement() returns nullptr?
There was a problem hiding this comment.
I did not change the already implemented behaviour. What do you think? Opinions of other developers are also wellcome. If we decide to change the behaviour should we do this in a separate PR?
There was a problem hiding this comment.
as i mentioned in my comments there are multiple ways to break this read-method. however, in my opinion we an leave it as it is for the most part as it is "only" a legacy interface and ogs5-read methods are similarly buggy (i.e. it is unlikely that someone comes up with a msh-file that will read in ogs5 but not in ogs6).
that said, in this particular case i would also not insert a nullptr but instead either abort reading the mesh or simply skip the element completely. in both cases we can guarantee consistent behaviour afterwards, whereas a nullptr within the element-vector will result in a crash later on...
There was a problem hiding this comment.
Changed as requested.
|
👍 |
TomFischer
force-pushed
the
MeshIOUsingNewPropertiesImpl
branch
from
9897935
to
e49195f
Compare
MeshIO using new Properties and PropertyVector.
|
OpenGeoSys development has been moved to GitLab. |
Reading material ids of elements from the old ogs mesh file format (*.msh) into the new data structures.