On making an AST about prog8 HOT 5 CLOSED

"a little writeup" 😄

from prog8.

Oh well...

from prog8.

@meisl a little history: I read a few articles by Federico Tomassetti about antlr and kotlin ast's, I think one of them was https://tomassetti.me/parse-tree-abstract-syntax-tree/ and that put me on the way of the 'toAst()' extension methods. They've since stuck as the mechanism to get the kotlin ast classes from the antlr syntax tree.

Finally, there's this https://github.com/Strumenta/antlr-kotlin/ but it's still experimental . It could perhaps be used to remove the antlr java parser code and replace it with kotlin parser code directly -- but we still want to convert the antlr syntax tree to our own specialized ast classes.

from prog8.

• IStringEncoding is no longer part of the CompilerAst (actually, in 7.2, it has now been moved to a new module containing a bunch of compiler related interfaces that are shared over other modules. CompilerAst is NOT among those btw)
• the containing module is not needed as global state to get the position from- these are extracted from the stream that the parser is reading from

from prog8.

while the complex "compiler Ast" as discussed above is still generated from the antlr parser, there is now also a new lightweight "codeAst" that is meant to be used in the codegenerator (this still has to be built)
So the idea is that the compilerAst is going to be internal to the compiler and optimizer phases and is no longer needed/visible in the codegenerators.

The new simpler Ast can be constructed incrementally by adding child nodes to incomplete parent nodes.

The compiler right now and the optimizers have way too much dependencies on the old Ast classes still, so replacing that by the new Ast completely is a bit too much for now, but maybe this can be done in the future to simplify things there as well.

from prog8.