Add "no-backline" feature

[GuillaumeGomez]

This is part of an effort to reduce the number of unneeded characters in the generated HTML. So using this feature when generating the std gave this result:

Using `du -s doc/std/`:

before: 105860
after: 105812

Using `wc -c std/index.html`:

before: 62859
after: 62631

So surprisingly, the number of characters seems to be quite high on a given file, but the impact on the documentation seems quite small.

So what do you think of this feature? Would you be interested into it? It diverges from the commonmark spec, hence why I suggest to add it as a feature and not a parameter.

[raphlinus]

Are these allocating? If so, that seems like a performance hit.

[GuillaumeGomez]

It's const string cancatenation so it should have no impact at runtime (to be confirmed though!).

[GuillaumeGomez]

Ah maybe not. I think String::from isn't const.

[GuillaumeGomez]

Originally I wanted to use if cfg!(feature = "no-backline") but found it too noisy. Well in any case, it's another possibility. But the question here is really if you're interested by this feature or not. If you are, I need to add tests too. :)

[notriddle]

I don't think LLVM is smart enough to get rid of the allocation.

Instead, you could make it with concat!? Like this?

#[cfg(feature = "no-backline")]
macro_rules! opt_backline { ($lit:lit) => { concat!($lit) } }
#[cfg(not(feature = "no-backline"))]
macro_rules! opt_backline { ($lit:lit) => { concat!($lit, "\n") } }

Then this line becomes:

Suggested change

	self.write(&(String::from("<br />") + BACKLINE))?;
	self.write(opt_backline!("<br />"))?;

[GuillaumeGomez]

Possibly. But considering the small impact, I think it's not worth it currently. Maybe for later on.

[marcusklaas]

I am not sure whether an output size savings of 0.05% to 0.3% is worth the extra complexity this would introduce, especially when we would have to use a bunch of cfg!(feature = "no-backline") checks to prevent performance regressions.

[GuillaumeGomez]

Considering the code it adds compared to the result, I don't think it's worth it currently so closing it.