Remove hls-tactics-plugin

[michaelpj]

While it remains very popular and desired, we don't have a viable maintenance path now @isovector isn't maintaining it and it doesn't work on the last three major GHC versions. @santiweight has made some attempts, but I'm not sure whether they're going to succeed.

At this point I think we should maybe just declare it dead.

[bolt12]

Personally I don't use all its features, I find the most useful ones to be just pattern matching and friends. Perhaps a good middle ground would be to strip it down to only the most used features? Hopefully that's easier to maintain

[michaelpj]

Maybe! And I believe there is some prior work in that direction. But someone has to take it on.

[isovector]

It's the pattern matching that is the maintenance burden, unfortunately. Everything else is more-or-less self-contained pure code. But spitting out the actual characters to change the document is excruciatingly hard.

[santiweight]

I am no longer able to push forward on this effort :(

I think that developer tooling is one of the biggest potential wins for Haskell, but unfortunately as Sandy says, the pretty printing infrastructure is fairly poor (very powerful, but extremely hard to use).

I think this is a good place for the Haskell Foundation to get involved.

[michaelpj]

Done 😿

[alanz]

@santiweight can we have a discussion about the pretty printing infrastructure some time?

[santiweight]

@santiweight can we have a discussion about the pretty printing infrastructure some time?

Would love to! Can you send me an email (we’ve exchanged before)?

[isovector]

@alanz @santiweight I'd love to be involved in that!

[alanz]

@isovector As I understand it the plan moving forward is to wait for my "simplify exact print in GHC" branch to land, and then see how we can clean up and improve the user-facing API.

I anticipate this taking around a month

[s-and-witch]

@santiweight @alanz there also is discussion in GHC GitLab about how we should store tokens in the AST, and it seems to be closely related to exact printing API, isn't it?
I think that it's good to know the opinion of HLS team about which plan is better.