We now have a lot of null values in layers, like in this example. The yaml files are already quite a bottleneck in terms of performance, this certainly will make the problem worse. Before we do more transformations into opf we really need to find a way to fix this issue.
I think a good solution would be to use yaml representers (or something similar) so that a pedantic class can be serialized to yaml and a yaml file can be read as a pydantic class directly (with no round trip with dictionaries)
Once I am done with current ongoing project. I will ask NT to set aside time to work on pydantic refactor and hopefully we will be able to fix this issue all together.
We now have a lot of
null
values in layers, like in this example. The yaml files are already quite a bottleneck in terms of performance, this certainly will make the problem worse. Before we do more transformations into opf we really need to find a way to fix this issue.This is linked to https://github.com/OpenPecha/Toolkit/issues/186 (I think) and it's become more apparent in a recent change (I'm not sure which one).
I think a good solution would be to use yaml representers (or something similar) so that a pedantic class can be serialized to yaml and a yaml file can be read as a pydantic class directly (with no round trip with dictionaries)