Field mapping for data lineage

I like the idea. Am I assuming you would like to see which column relates to a different column in a different table?

If not, I would like to expand to show the exact relations between tables i.e. when ex1.UID = ex2.UID 

Hi,

I do not know how they do their lineage - if it is AI/LLM driven it would need to be very good to convince me to trust it. If syntax analysis is not 100% factual, it is useless in my opinion. Given that language is ambiguous and the object of code is to be precise, LLMs are not a natural tool to solve such problems.

From the sparse documentation they seem to analyse logs among other approaches. The documentation does specify things like: "Sometimes lineage may be missing on a table (or other item). To update the Lineage information, first go to the Data Item’s Profile Page” which tells me there will be a lot of user interaction required similar to Purview's lineage extraction to get complete coverage. 

I have seen some other tools that generate good-ish lineage if there is user input to give the tool enough guardrails to generate something, but there is usually a big trade-off involved. 

There is a wider discussion underlying on whether a percentage less than 100% is good enough to be workable. In my opinion in this case not, but I suspect the LLM-era will try to move this bar to a lower number. After all, if you are tracing a bug, you don't want to spend time chasing a red-herring caused by a guesstimate. Similarly to prove compliance with regulation, you need a perfect analysis to avoid costly legal issues later.

To be honest It would be great if there was an auto parameterization in TX

Where it checks the names in the project with the names of the tables and then adds the columns in the table script to the parameters.

or when you drag a table to the inputbox it automatically adds (when possible) all the columns to the parameters (because when you exits TX can remove the unused parameters automatically).
Wold be super nice if TX add the parameters automatically and in such a way that the names are always unique per table so everything get 100% good parameterized!