Experimental implementation of a new parser for RON using small functional parsers for individual syntax elements over a stateful parser.
This experiment started off as "RON + nom", but is now using its own
parsers & combinators inspired by nom.
The current ron parsing suffers from the following problems:
- parsing is done in both
parse.rsand the deserializer itself- bad code organization
- hard to maintain
- limits reuse
- serde's data model stops us from accurately reflecting struct / map and struct names
Stateless / functional parsers are
- easier to maintain
- easier to reuse
- much easier to test
An abstract syntax tree (AST)...
- makes the deserializer much easier & cleaner to implement
- allows reporting locations of syntax & type errors
- can be reused by multiple deserializer implementations (
serde::Deserializer,our_own::Deserializer,ron-edit)
The old RON deserializer produced errors that were often hard to understand.
ron-reboot is meant to change that; this is the output of a deserialization
error as of 2021-10-15:
error: invalid type: boolean `true`, expected a string
--> string:3:9
|
3 | y: true,
| ^^^^
|
or with a multi-line expression:
error: invalid type: map, expected a string
--> string:3:9
|
3 | y: (
| ________^
4 | | this: "is",
5 | | not: "the right type",
6 | | ),
| |______^
|
| Goal | Status |
|---|---|
| Parser generating AST | ✔️ done |
| Parser generating beautiful errors | ✔️ done |
| Accurate Value deserialization from AST | ✔️ done |
| Replace nom combinators | ✔️ done |
| Spans in AST (locations for error reporting) | ✔️ done |
| Serde Deserializer using AST | ✔️ done |
| Serde Deserializer generating beautiful errors | ✔️ done |
ron-edit (format & comments preserving writer) |
❌ to be done |
Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, shall be licensed according to the license of this project (Apache-2.0), without any additional terms or conditions.