Hi all,

I've had a busy past week figuring out the next steps in the dfmt rewrite:

• The lexer has been completely replaced across the utility functions, and dfmt compiles with dmd! Figuring out the Makefile changes for the build took some fiddling, but I got it working in the end.
• The parser from libdparse has been removed, and dfmt currently uses the AST from dmd. At the moment, the AST is successfully built, but no transformation passes are available.
• Going forward, I will be working on getting one of the existing passes working with the dmd AST. If one works, then the rest can be ported pretty easily.
• Tangentially, I realised that dfmt does not perform any semantic analysis or use any of the data that the AST provides. Validating the correctness of the code being formatted (whether it can generate a valid AST or not) is not the responsibility of the formatter. So, I'll also be trying some experiments to see if we can perform the formatting passes purely using the token stream from the lexer. There is a chance we might not need a parser at all!

On Friday, 29 September 2023 at 06:48:28 UTC, Prajwal S N wrote:

That would nice. Good luck.