some suggestion

[SherlockNovitch]

Hi, after reading your documents and source code, I feel that the interface design is a bit unfriendly. Sometimes, users only need to parse a certain section of code or specify a single file for parsing, but SOLP requires the entire project to be input. Users may need an interface style similar to tree-sitter.

[t81lal]

Hey sorry I didn't see this before.

You can parse a few files on their own but if the code has references that can't be resolved properly at compiletime, you'll only be able to get an AST1 output: it's detailed in this tutorial (up to the Getting AST1 Nodes section).
I think you're talking about the boilerplate required by setting up a VFS though and I agree it's quite clunky. Do you have an idea for how you'd want this interface to look like?

[Ruschio]

@t81lal Is it possible to convert AST1 to AST2 format? If yes, is it possible also if there are some missing dependencies?

[t81lal]

@Ruschio Usually you would setup a VirtualFileSystem environment with all your source code/libraries, create symbol tables, then feed that into AST2Builder: https://github.com/Zellic/solidity-parser/blob/main/example/quickstart.py#L43

If you have some top level AST1 nodes that you've already manipulated and want to turn into AST2, you could try to manually queue them into the AST2Builder: https://github.com/Zellic/solidity-parser/blob/main/src/solidity_parser/ast/ast2builder.py#L943 (n is the AST1Node that will be converted).

As far as missing dependencies: AST2Builder needs to be able to resolve the a file's dependencies transitively to be able to load that file. If you have code that is missing dependencies, you can comment out parts that link to that dependency.

[Ruschio]

@t81lal Thanks for your answer!

Anyway, I'm trying to generate AST programmatically from an input path and handle them in a common way (AST1 or AST2), thus I can't know in advance if the contract contains missing dependencies. So, since it's not possible to parse from AST1 to AST2 if there are missing dependencies, can I parse from AST2 to AST1 and handle either contracts with missing dependencies and complete ones? Is there advantages in translating AST2 back to AST1 or it will contain the same information as I parse the contract directly with AST1?

[t81lal]

AST1 was designed to let the client read and modify Solidity code in a version agnostic manner without complete dependency resolution. AST2 is meant to provide a better API experience by linking different AST2 subtrees(e.g. state variable load expressions link to the state variable node that it's loading and free floating function definitions get put into synthetic file definition nodes so the nodes always have a parent attribute, stuff like that) and there's currently no way for the builder to handle missing links.

There isn't a 1 to 1 mapping of AST2 to AST1 nodes but iirc there are some AST2 nodes that have an ast1_node attribute on them (I think top level nodes). I wouldn't use that however, if your tool needs to handle missing dependencies, I'd just build the analyses using AST1: AST2 only provides a slightly better developer experience and your analyses can load AST1 types manually on demand instead of needing AST2 to link them for you.

Hope this helps, feel free to email me(bilal@zellic.io) if want to share some code or get help on something you can't post publicly here.