[web-devel] [hamlet] implicit spaces with newlines?

[Michael Snoyman]

On Thu, May 19, 2011 at 11:49 PM, Patrick Palka <patrick at parcs.ath.cx> wrote:
> I find it a bit unintuitive that the hamlet code
>
> <p>hello
>    <strong>there
>
> or
>
> <p>
>     hello
>     <strong>there
>
> generates the html
>
> <p>hello<strong>there</strong></p>
>
> I expected there to be a space between "hello" and "there" similar to what
> the html specifications dictate. Is this behavior intentional or an
> oversight? If it's the former, then what is the recommended way to simulate
> my expected behavior? Appending a space to the end of a line is
> mentally ugly and syntactically obscure.

I'm not sure what you mean by "what the html specifications dictate."
HTML is whitespace-sensitive, meaning that:

   <i>foo</i> <b>bar</b>

and

    <i>foo</i><b>bar</b>

Are different. Now, in all likelihood in the above example, you will
want to have the whitespace surrounding tags. But consider the
following HTML:

    <p>You are logged in as <i>Michael Snoyman</i>, <a
href="/logout">logout</a>.</p><p>Another paragraph.</p>

In the case of the <i> and <a> tags, we definitely do *not* want to
add whitespace after the tag (though we do want it before the tag). In
the case of <p>, we don't care one way or another, but adding the
whitespace everywhere will take up (a trivial amount of) extra
bandwidth. tl;dr: Sometimes you don't want the whitespace.

So when designing Hamlet, I thought up a few possibilities:

1) What we do now: all whitespace must be explicit.
2) Implicitly add whitespace before/after every tag.
3) Do something "smart", adding whitespace where it's desired.

(2) isn't really an option because it makes having a tag as the last
word in a sentence impossible. (3) gives me the creeps: I like smart
libraries, but I will *never* trust a library to do this kind of stuff
correctly all the time, even if I'm the one making up the rules for it
to follow! And I have no doubt that it will quickly devolve into 500
lines of hairy code to try and cover millions of corner cases. Oh, and
don't forget that there are other languages than English that might
approach it differently.

I suppose another possibility is (2) along with some special way of
forcing the removal of extra whitespace, but this seemed much less
intuitive than the current approach.

Anyway, that's the reasoning behind this stuff, if people have better
ideas, I'd like to hear them.

Michael