WIP: Make Ident a Functor

[kalhauge]

Hi all,

Thanks for all the hard work you put into this package.
I was currently working on a project that produced C files, and I want to test that the Ast I create can be read again. However with the current setup it is very hard to compare two AST's.

This is only meant for inspiration, and should not be merged as is.

What I have done:

• Made Ident a functor
• Created data types for all tuples used in the AST (which made the AST not able to derive Functor)
• Made the AST able to check if it is equal with another AST.

[expipiplus1]

I wonder if being polymorphic over the NodeInfo it would be worth removing the weirdy Eq/Ord instance. One could always do something like (==) `on` void to regain the old behavior.

[kalhauge]

Yes! great catch! I think it would make sense to stay with the stock implementations.

[expipiplus1]

Thanks! I'm all for this. What's left to do before dropping the WIP flag?

[kalhauge]

Thanks! I'm all for this. What's left to do before dropping the WIP flag?

Happy you like it :). The primary reason it is WIP is because I did not know if it was in the spirit of the repository, but now that you approve with the direction I think that what is missing, considering this is a already a breaking change, is:

• Add documentation for the newly added data types,
• Make sure the tests are succeeding,

Further more, I think that we could improve the code a little by, but that might fit better in another pull request:

• Be consistent with the use of Generic1,
• Maybe let Annotated be generated using Generic,
• CNode is now just nodeInfo = annotation for all classes,
• Pos is always posOf = posOf . nodeInfo, and finally
• CDeclarationItem Seems to only have four valid 'states', however the current (Maybe a) (Maybe b) (Maybe c) pattern allows for 8 'states'. Maybe making those 'states' explicit might make the code easier to read (I have started with the CDeclI pattern).