Join GitHub today
GitHub is home to over 31 million developers working together to host and review code, manage projects, and build software together.Sign up
Follow up to #1559 : have the name of unknown missing tag depend on the row #1564
trefis left a comment
To clarify: I was fine with the state of things after the last PR got merged, the emitted warnings were less pretty, but correct.
The current GPR makes the warnings pretty again in many instances.
As for the implementation: it seems correct to me.
I prefer occasional clarity to consistent strangeness. (But I would still vote for AnyOtherTag rather than AnyExtraTag.) Furthermore, I think this PR highlights and enables an interesting distinction (private row variables) that suggests an explanation in this specific difficult case (we could have a mention in the manual that "" shows up for private rows and explain why): we can do better thanks to the distinction.
Small remark: the algorithm is not quadratic in the number of discriminating constructors, it is just < the number of discriminating constructors > * (< number of `AnyOtherTag among them > + 1). The last one is 0 in 99.99% of cases, and at most one except if you are intentionally trying to fool the algorithm. So this should be good enough.
Concerning documenting the change, where should it go? I'm OK to adding something to the warning section of the manual, but this already relatively self-explenatory.