Grammar does not support CPP Directives about tree-sitter-haskell HOT 6 CLOSED

Rather than "C++", do you mean CPP, the C Pre-processor?

from tree-sitter-haskell.

@maxbrunsfeld Thanks, I've updated the original post.

from tree-sitter-haskell.

while parsing doesn't fail on #ifdef et al anymore, it still cannot deal with #else other than creating a single node for the entire code block.
The reason for this is a general problem, tree-sitter's state would need to be reset to what it was at #if when encountering #else, and even then I'm not sure how the parse tree would look like.

Another problem with CPP is that macros can be used anywhere and contain anything, causing potentially impossible situations.
There's this case in polysemy:

#if __GLASGOW_HASKELL__ < 806   -- There is a bug in older versions of Haddock that don't allow documentation on GADT arguments.                                                                                                   
#define HADDOCK --                                                                                                      
#else                                                       
#define HADDOCK -- ^                                                                                                    
#endif  

so the comment marker is inserted by the macro.

I can't imagine how this would be handled, even if we would run the preprocessor over the file, the positions would be off.

from tree-sitter-haskell.

There would need to be an API function available to the scanner that allows to inject shadow code into the source that is parsed regularly and mapped back to the original positions.

from tree-sitter-haskell.

Preprocessor-based parsing will always be a good-enough effort (modulo products like CodeQL that attempt to recreate the build environment in toto). I think we should keep this open for informational purposes, but also not sweat it too much.

from tree-sitter-haskell.

sounds good.

from tree-sitter-haskell.