Whitespace in HTML macro

[dbrgn]

How do I create I the equivalent of Hello <img src="#" alt="Big"> World with the html macro?

When I do html! { <div> Hello <img src="#" alt="Big"> World </div> } then the whitespace around the text is stripped. Can it somehow be escaped without using the text! macro and variable replacement?

[chinedufn]

Here's how we're able to parse text without needing quotation marks

percy/crates/html-macro/src/tag.rs

Line 75 in 0af49f7

return parse_text_node(&mut input);

Basically if we come across a token and it doesn't look like anything else (a tag.. attributes.. etc) then we assume it is text.

---

When parsing tokens from the compiler we have no idea how much spacing was around the token. We just get raw tokens.

So when we're building a string of text we just insert space between tokens ourselves:

percy/crates/html-macro/src/tag.rs

Lines 202 to 204 in 0af49f7

	if idx != 0 {
	text += " ";
	}

---

As you see in that code we are only inserting a space after a token if it isn't the first text token in the batch of text.

In your example above Hello and World are both the first (and only) text tokens in the batch of text tokens that we see so neither of them is getting any space in front of them.

---

One thing that would really help here is being able to know the start and end columns of each token

https://docs.rs/proc-macro2/0.4.24/proc_macro2/struct.Span.html#method.start

This way we could look at the two neighboring tokens and know whether or not we need to insert a space character.

Unfortunately this feature of spans hasn't stabilized and I'm not sure if/when it will.

---

Potential options in the meantime...

1. Always insert space before and after text tokens. This would mean that if you really didn't want a space between an image and some text after it there would be .a space anyway... but maybe that's a trade-off that we can mark for now.

2. Add support for literals html! { <div>"This will properly be parsed "</div> }

• I want to do this regardless because of Force quoting punctuation #69

3. Remove support for un-quoted text until span starts / ends stabilize

• fewer surprises but also much less ergonomic IMO. I don't like this and would rather have this bit of sometimes unexpected (on the first and last word in a sequence of text) behavior than throw away the ergonomics here.

---

Your thoughts?

[dbrgn]

Supporting literals sounds like a good idea!

I'm not sure about un-quoted text to be honest. It's strange to have two ways to declare text nodes, but on the other hand I think it's fine as long as the differences are documented (i.e. "white-space is collapsed").

[chinedufn]

Hmm I'm thinking that ideally we'd support all ways that someone might try to create things.

So the problem here is less about having two ways and more about having a way that might behave in a way that you don't expect sometimes.

(As always, just my thoughts, very open to being wrong!!)

I can try to figure out where we can follow the progress of stabilization and in the meantime add literal support.

Let me know if you have something different in mind!

[chinedufn]

Woah.. wait a second.. this might be possible right now. dtolnay/proc-macro2#166

Doing some reading ...

[chinedufn]

@dbrgn started working on this - should have a PR ready this week if you can take a look

[chinedufn]

Alright all of the test cases are passing and can be seen here

https://github.com/chinedufn/percy/blob/text-space/crates/html-macro-test/src/text.rs

Will need to do some refactoring / clean up of the html-macro and then we'll be good to go!

[dbrgn]

@chinedufn fantastic, exactly what I'd intuitively expect! :)