Related #366.

Thanks for tackling this one. This is important feature.

Having said that I think it should be done a indentation style setting. Either tab-width is important or alignment is important, not both.

Rationale:

Indentation in Haskell is mostly deterministic except in a couple of places when a layout block ends. Keeping it as deterministic as possible will enable us to have a sane indent-region.

Your commit as it is now would increase non-determinism, which is a very sad trade off to take.

This patch is very stupid and code in haskell-indentation.el is too complicated to understand it in measurable time. I think this patch is the best I can offer.

Or you just mean that I must use haskell-indentation-layout-offset (or some other setting) instead of tab-width ?

Your patch is an important vote for shallow indentation. Thanks.

There is a place in the code where indentation offset for the { is calculated. Can you find it?

Well, there is many places where { pattern occures in code. haskell-indentation.el has very very complex code. To do this patch I just grepped possible-indentations defined in haskell-indentation-parse-to-indentations assuming it is a list of posible indentation positions. I found one such place, which is function haskell-indentation-push-indentation, then I modified this function and trick worked.

Code in haskell-indentation.el is not very functional-style, everything mutates everything, it is very complex to understand. We need someone who knows how all this works to make this feature right.

That is why current project focus is lowering the entry barrier for new users, new issue reporters and first time contributors and the goal is 100 haskell-mode contributors.

We need to leave after us a maintainable piece of software. I've started doing this by writing some unit tests like in #562.

While you tried to hack on haskell-indentation you have learned a lot already. Can you document your knowledge in the form ofo unit tests, comments or refactorings?

Well, I will try, but can not promise anything usefull.

I'm closing this PR with strong conviction that #366 is the way to go.