refactor - move from macro to shapeless

[kailuowang]

The incentive is rather philosophical
1, macro should remain low-level and exists at the lower level libraries such as shapeless.
2, we should use each other's libraries more.

[fommil]

happy to help you write this! My talk at Scala eXchange last week should make this real easy. I predict you could do the whole thing in about 50 lines of code, and also support sealed traits.

[kailuowang]

That will be fantastic! I bet you can write it much better than us.

[fommil]

well I didn't say I'd write it, but I'd be happy to show you how 😄 When my shapeless talk is out, I'll link you and then give you guidance on the example, then you can graduate to rewriting ficus. Ping me in a few weeks if I don't come back.

[kailuowang]

That's still very nice of you. Looking forward to your talk!

[fommil]

enjoy! https://twitter.com/fommil/status/678195060429705216

[kailuowang]

Thank you @fommil , I enjoyed your talk. The explanation of that several weird things shapeless has to do to overcome Scala compiler's defects can be immensely helpful for new comers to typelevel projects. I suggest shapeless have a section on these issues in documentation.

[kailuowang]

A note from ceedubs,
Travis Brown has done some recent work to support different cases in Circe. Since it uses Shapeless for derivation, it might be a good reference.

[fommil]

yup, but the shapeless for mortals talk is still probably a better learning experience. Travis was in the front row, btw 😉

[fommil]

any movement on this?

[kailuowang]

@fommil I made some progress but not there yet. I am trying to build a library that can "read" any runtime structured data (Map[String, Any], Typesafe Config, Json AST, etc) into any case classes. I am still couple of days away from knowing if I can pull that one out yet.

[fommil]

@kailuowang I hope you've watched the talk ? Please don't make this library do any more than typesafe's config. There are already libraries to do the others.

[kailuowang]

Sam, I watched the talk. I am don't plan to make this library doing any
more than typesafe config. The thing I was talking about is a different
project.

On Wed, Jan 27, 2016 at 10:18 AM Sam Halliday [email protected]
wrote:

@kailuowang https://github.com/kailuowang I hope you've watched the
talk ? Please don't make this library do any more than typesafe's config.
There are already libraries to do the others.

—
Reply to this email directly or view it on GitHub
#5 (comment).

[fommil]

cool! Hope it helped 😄

[andr83]

May be if it is interesting I did simple implementation on shapeless: https://github.com/andr83/scalaconfig

[joprice]

This library has a slightly wider scope than product/coproduct hierarchies. Trait, class, and enumeration support may have to be dropped or remain as is. Also, if compilation speed is affected, I don't see any immediate benefit for users.

[andr83]

I thought this issue discuss refactoring to shapeless and decide to share my implementation. Regarding this do not see any compilation speed trouble before start use case classes with hundreds of fields. Yes my solution do not support traits with apply factories and enumeration but classes and custom readers are supporting. If missing features important for someone I will implement it.

[joprice]

I didn't meant to target your comment. It simply brought the issue to my attention again, and I was voicing my opinion on the general topic.

[fommil]

@andr83 great! What license are you going to use?

[andr83]

@joprice sorry for my confusion - our comments were close by time after a big delay in this topic )
@fommil it's publishing by MIT license. You can see it on project github page. Any suggestions are welcome )

[fommil]

Apache 2.0 is better than MIT because it covers a lot of things MIT never considered but you almost certainly want if you want to go the permissive route. If you want the software to always be libre, then I recommend MPL or LGPL. I did a talk about this somewhere that you could find by googling my name and scalasphere

[andr83]

@fommil I'm agree with you on 100%. But exactly about this library I have no worries for any usage so MIT fit it well. Of course for other projects which I want to protect I will use more suitable license.

[fommil]

Still, Apache 2.0 is an upgrade on MIT and is the license preferred by shapeless :-)