feat: support for custom node

[ModProg]

fixes #11

[ModProg]

@vldm

I added parse_custom next to parse_simple and made parse_simple use Infallible, sadly type defaults are not supported on functions. It might make sense to do similar for other functions in order to avoid having to specify the generic.

Trying to implement a CustomNode for this syntax as a test, I noticed that we should probably expose more helpers to help with rstml syntax:

#[derive(Debug, syn_derive::ToTokens)]
struct If {
    token_lt: Token![<],
    token_if: Token![if],
    condition: Expr,
    token_gt: Token![>],
    #[to_tokens(|tokens, val| tokens.append_all(val))]
    children: Vec<Node>,
    close_tag: atoms::CloseTag,
}
html! {
  <if condition>
      children
  </if>
};

condition should use the same parsing, OpenTag::attributes have, i.e., first split the input at > and then parse it.

children should use the same parsing as NodeElement::children, and preferable NodeElement::close_tag as well.

My idea was to add parse_simple_with_ending and parse_node_body to RecoverableContext to allow this:

fn parse_element(
    parser: &mut rstml::recoverable::RecoverableContext,
    input: syn::parse::ParseStream,
) -> Option<Self> {
    let token_lt = parser.parse_simple(input)?;
    let token_if = parser.parse_simple(input)?;
    let (condition, open_tag_end): (_, OpenTagEnd) = parser.parse_simple_with_ending(input)?;
    let (body, close_tag) = if open_tag_end.token_solidus.is_some() {
        // Passed to allow parsing of close_tag
        parser.parse_node_body(&OpenTag {
            token_lt,
            name: parse_quote!(#token_if),
            generics: Default::default(),
            attributes: Default::default(),
            end_tag: open_tag_end.clone(),
        })?
    } else {
        (Vec::new(), None)
    };
    Some(Self {
        token_lt,
        token_if,
        condition,
        open_tag_end,
        body,
        close_tag,
    })
}

[ModProg]

@vldm the ci failed due to a case change in tarpaulin AFAICT, I made a separate pr for that #36, and rebased this one

[vldm]

@ModProg Sorry, was busy in road trip. I will review pr today.

[vldm]

Awesome work, thanks for contribution.

[ModProg]

For example, using parse_simple_with_ending in If prevent user from using > (great sign, compare operator) in condition, but allows to use < which i find controversial.

This is also true for attributes, though, isn't it? It should be documented, but that seems like a weakness of the syntax in general, not just of this usage (though comparisons are probably more likely in an if than in an attribute).

[vldm]

For example, using parse_simple_with_ending in If prevent user from using > (great sign, compare operator) in condition, but allows to use < which i find controversial.

This is also true for attributes, though, isn't it? It should be documented, but that seems like a weakness of the syntax in general, not just of this usage (though comparisons are probably more likely in an if than in an attribute).

Yes, this is true for attributes, and yes this should be documented.
But my point was about making methods public, not if implementation.

[ModProg]

One consideration, currently I'm using Infallible as the pseudo-! type.

But one problem is, that that means, we need the custom to_tokens function in CustomNode and cannot use the ToTokens trait, as Infallible doesn't implement ToTokens.

We could instead use our own Never type to work around that, or keep the additional trait method as is..

[codecov-commenter]

Codecov Report

Merging #34 (f2be1ef) into main (aa9d210) will decrease coverage by 0.09%.
The diff coverage is 86.30%.

❗ Current head f2be1ef differs from pull request most recent head 1907a7a. Consider uploading reports for the commit 1907a7a to get more accurate results

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

@@            Coverage Diff             @@
##             main      #34      +/-   ##
==========================================
- Coverage   77.31%   77.23%   -0.09%     
==========================================
  Files          11       12       +1     
  Lines         692      751      +59     
==========================================
+ Hits          535      580      +45     
- Misses        157      171      +14     

Files	Coverage Δ
src/config.rs	73.21% <100.00%> (+11.05%)	⬆️
src/lib.rs	42.85% <100.00%> (+17.85%)	⬆️
src/node/raw_text.rs	93.75% <100.00%> (-2.00%)	⬇️
src/parser/mod.rs	70.00% <100.00%> (-14.85%)	⬇️
src/parser/recoverable.rs	70.00% <100.00%> (+12.85%)	⬆️
src/node/mod.rs	60.86% <60.00%> (-4.99%)	⬇️
src/node/parser_ext.rs	80.00% <80.00%> (ø)
src/node/parse.rs	86.52% <82.14%> (-3.35%)	⬇️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[vldm]

One consideration, currently I'm using Infallible as the pseudo-! type.

But one problem is, that that means, we need the custom to_tokens function in CustomNode and cannot use the ToTokens trait, as Infallible doesn't implement ToTokens.

We could instead use our own Never type to work around that, or keep the additional trait method as is..

I am think current approach with custom trait and Infallible is good enough.

I‘d rather have adapter (wrapper type, or macro) that will derive implementation from ToTokens/Parse. For those who want implementation of this trait effortlessly.

[vldm]

Awesome!