Figure out a testing story. about bikeshed-rs

[annevk]

DOM-based comparison seems a little more flexible so you can document in code what kind of whitespace or attribute differences are acceptable. If you do string-based you might run into issues such as the hashmap ordering of attributes being different.

[annevk]

See https://github.com/tabatkins/bikeshed/tree/master/tests. The idea is, given an input .bs, does it generate the expected .html.

[whichxjy]

I use the serialize method of kuchiki to serialize the DOM tree now, but there's a special HTML serializer in bikeshed. If we want to use the expected HTML files in bikeshed for testing, maybe we need to write a serializer like that.

[emilio]

Alternatively, can we parse the bikeshed HTML with kuchiki, and serialize it back to guarantee it's idempotent?

Output doesn't have to be whitespace-compatible IMO.

[whichxjy]

That makes sense if there's nothing special about the serializer in bikeshed. What about building DOM trees from source files and expected HTML files, and then comparing them without serialization?

[annevk]

That's also fine, but you'll still run into the whitespace issues I suspect. And potentially attribute ordering issues as well. (Attribute order is unfortunately still implementation-defined.)

[emilio]

Yeah, however those issues should be easy to work around / ignore in the comparison (sorting the attributes, and skipping child whitespace text nodes where appropriate).

So it seems pretty reasonable to me.

[whichxjy]

I'm trying to compare DOM trees in this way:

fn is_equal(lhs: &NodeRef, rhs: &NodeRef) -> bool {
    if lhs.data() != rhs.data() {
        return false;
    }

    let lhs_children = lhs.children().collect::<Vec<NodeRef>>();
    let rhs_children = rhs.children().collect::<Vec<NodeRef>>();

    if lhs_children.len() != rhs_children.len() {
        return false;
    }

    for (lc, rc) in lhs_children.iter().zip(rhs_children.iter()) {
        if !is_equal(lc, rc) {
            return false;
        }
    }

    true
}

But it isn't working because whitespace characters that are outside of HTML elements will be treated as text nodes (kuchiki will treat newline character as NodeRef(Text(RefCell { value: "\n" })).

• If we want to do the DOM-based comparison, we need to handle these text nodes in DOM trees.

• If we want to do the string-based comparison, we need to ensure the generated HTML files and the expected HTML files have the same format.

[whichxjy]

The DOM-based comparison is working now. See #12.