Allow HEAD requests to be served by fs::dir() filter

[zenria]

This is a fix for feature request #834

[joseluisq]

How about the empty body for HEAD responses? https://tools.ietf.org/html/rfc7231#section-4.3.2

[zenria]

The underlying HTTP server (not sure if it's hyper or tower) does not sent the body when handling HEAD requests.

This has already been done in the fs::file() filter:

warp/src/filters/fs.rs

Lines 48 to 57 in 4a03710

	pub fn file(path: impl Into<PathBuf>) -> impl FilterClone<Extract = One<File>, Error = Rejection> {
	let path = Arc::new(path.into());
	crate::any()
	.map(move || {
	tracing::trace!("file: {:?}", path);
	ArcPath(path.clone())
	})
	.and(conditionals())
	.and_then(|path, conditionals| file_reply(path, conditionals))
	}

[joseluisq]

Great.
Another question does the file filter need HEAD method support too or not?

[zenria]

I think, yes it needs it, but this is already working due to the any() filter.

I didn't want to use the any() filter for the fs:dir() as it appears cumbersome to respond to POST or PUT requests on static files. So I think we should also disable this on the fs:file() filter and use only get+head filter

[joseluisq]

I think, yes it needs it, but this is already working due to the any() filter.

Do you mean this is working for your addition or do you mean this file filter is already working that's why you don't want to break someone else code?

But yeah I'm also wondering in what cases we want to support fs::file() with any method.

[zenria]

I think, yes it needs it, but this is already working due to the any() filter.

Do you mean this is working for your addition or do you mean this file filter is already working that's why you don't want to break someone else code?

the second option: I did not want to break someone else code!

But yeah I'm also wondering in what cases we want to support fs::file() with any method.

:-)

[jxs]

thanks @zenria! and thanks @joseluisq for the reviewing :)

The underlying HTTP server (not sure if it's hyper or tower) does not sent the body when handling HEAD requests.

yeah it's hyper, see here

[amousset]

FYI this change actually broke some code when updating from 0.3.1 to 0.3.2. We had a custom handler for HEAD but relied on fs::dir() for GET requests (something probably quite unusual I admit :) ). Adding a method::get() filter before the fs::dir() fixed the problem (and I think serving HEAD too makes more sense).