SourceCode on Option<impl SourceCode>?

[mimoo]

I would find it useful if there were such an implementation:

impl miette::SourceCode for Option<NamedSource> {
    fn read_span<'a>(
        &'a self,
        span: &miette::SourceSpan,
        context_lines_before: usize,
        context_lines_after: usize,
    ) -> std::result::Result<Box<dyn miette::SpanContents<'a> + 'a>, miette::MietteError> {
        self.expect("no source code was set in the error").read_span(span, context_lines_before, context_lines_after)     
    }
}

or perhaps a more general implementation:

impl<T> miette::SourceCode for Option<T> where T: miette::SourceCode {
    fn read_span<'a>(
        &'a self,
        span: &miette::SourceSpan,
        context_lines_before: usize,
        context_lines_after: usize,
    ) -> std::result::Result<Box<dyn miette::SpanContents<'a> + 'a>, miette::MietteError> {
        self.expect("no source code was set in the error").read_span(span, context_lines_before, context_lines_after)     
    }
}

this way I can have an optional source up until I set it:

#[source_code]
src: Option<NamedSource>

BTW I also noticed that I can't call this field source otherwise I get a weird error. Might be good to have a better error when I call the field source instead of src

[olivia-fl]

Having an implementation for Option<T: SourceCode> that panics at runtime when it's None is not ideal, since it makes it easy to accidentally cause a panic when trying to print an error. I think there's a better option though. The Diagnostic::source_code method returns Option<&dyn SourceCode>. Currently the Diagnostic derive macro always returns Some(src) when a field is annotated with #[source_code], and always returns None otherwise. We could update it to pass src through as an Option when applicable, which should handle your use-case.

[erratic-pattern]

Just wanted to say that I would also find this very useful on #[diagnostic_source] as well, to optionally provide more detailed debugging info based on environment variables, cli options, configuration, etc. This could also use a similar implementation to what @Benjamin-L suggested for #[source_code]

[zkat]

this might be a nice thing to have, yes.

And regarding calling the field source: this is because thiserror uses that field name to determine its own source. It's not a miette thing.