We have started working using PBIP and Git integrations recently. It's been helpful but what's frustrating is the long list of Diffs we see on TMDL files where the relationships / measures being reor...
I strongly support that this is a huge issue, if you have a semantic model with enough measures this becomes a constant pain and costs you a lot of time. However, I think the correct solution is not to change serialization behavior, but to change the design of TMDL completely. Honestly I am baffled at how anyone can think that putting all measures and columns in one file in a source code structure which specifically aims to enable management with git is a good idea. The ordering of measures and columns within a table OBVIOUSLY has no meaning! What makes it even worse is that this is a solved issue, the 3rd party library pbi-tools made the correct design choice and serialized each measure and column as individual files. Then, you can sort the files however you want, it won't leak into your repo.
