Ambguity in the Smithy IDL grammar

[hlbarber]

One example of an ambiguity in the grammar is within shape_section given in the Smithy IDL specification under "Smithy IDL ABNF" -> "Shapes".

shape_section : [`namespace_statement` [`use_section`] [`shape_statements`]]
use_section : *(`use_statement`)

Is [`use_section`] empty or is it populated with an empty *(`use_statement`)?

This places Smithy IDL outside of a large class of grammars (see page 4 of Miscellaneous Parsing).

An example of where this is a problem is when trying to implement the grammar in lalrpop which will complain when we try to implement the following

use_section = use_statement*;
pub shape_section = use_section?;

A similar conflict is described in this issue.

The ambiguity around [`use_section`] can be resolved by making it non-optional or changing use_section to 1*(`use_statement`).

• Should we attempt to resolve syntax ambiguities?
• Should we document how to resolve conflicts?
• Is it permissible to change the grammar if it does not change the language/semantics?
• Should we document the formal properties of the grammar?

Disclaimer: I'm not an expert in language theory, please be skeptical of my claims and correct me where needed.

[mtdowling]

I don't think this is ambiguous, but sure the [ and ] around use_section is unnecessary since *(use_statement) is already 0 or more.

And yes, we can change the grammar if it's a bug fix. Please feel free to send us a PR!