ohsh6o
commented
3 years ago I spent a good portion of the morning and afternoon looking at XML semantic diffing approaches. A lot of the algorithms appear to be complex and not easily implemented in XSLT itself (more experimentation is needed, but many of them utilized different forms of hashing and other constructs that are not offered in the language as defined in the XSLT 3.x spec).
I also investigated a number of tools for guided and/or automated semantic diffing (not line-based and/or text-based diffing) and none of them really worked. For now, it appears I will have to write some document guidance and tips and we continue to update that, or revisit some programmatic method at a later time, when resources permit.
In the past, we performed model changes inline in Github issues, but it might be best to document them more formally, in a document inside the code repository.
Example from ongoing issue in GH in comments that we converted into an epic now. Should move away from that and more formal and make it more accessible.