Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 95.18%. Comparing base (eb64722) to head (d4a3d79).

@@            Coverage Diff             @@
##           master     #903      +/-   ##
==========================================
+ Coverage   95.12%   95.18%   +0.05%     
==========================================
  Files           8        8              
  Lines        2196     2220      +24     
  Branches      577      585       +8     
==========================================
+ Hits         2089     2113      +24     
  Misses        107      107              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Sorry for the delay in processing this. I still have one very busy week in front of me, but afterwards I should be able to find the time for this topic.

Hey @hungarian-notation I finally managed to look into it closer and still have some doubts:

I'm still having a hard time wrapping my head around the topic, but I know that validateAndExtract, which is called as part of most if not all *AttributeNs menthos, prevents combinations of namespaceURI being defined but the qualified name not having a prefix. SO why doesn't it complain in this case?
I wonder if we can instead fix it in a way to make it impossible to produce the invalid DOM state?
Maybe we can have a call to help me understand all the details of it? If you are up for it, please reach out to me via e-mail from my GitHub profile.

PS: I'm not in reach of any computer and very likely also only with limited internet connection on the phone until 4th of August, so I will only be able to reply afterwards.

The rest of the message is only relevant, if those operations need to be possible and the fix can only ever be applied as part of serialization:

I think I detected some flaws in your implementation of assumePrefix:

• prefix defaults to an empty string, so there is no need to check for null later
• there is some redundancy in the early returns, I think some of them are not needed
• there were some == comparisons, we should always use ===
• let's not add module level variables that might never be used, I would prefer to define them inside the function and only once they are needed
• the detection of existing generated namespaces with all the slicing and NaN detection is very verbose and not ideal, since it also takes place in the case where the last visible namespace is an exact match and returns existingPrefix
• existing generated namespaces are only present when they have already been parsed, right? How do we ensure that we are not reusing generated prefixes across an element/node, since they are not added to visibelNamespaces, or are they?
• (minor) I would prefer the name serializableNamespacePrefix, does it also make sense for you?

I have refactored the method accordingly, started adding some basic tests which left me puzzled how to reproduce the different cases.
It would be great of you could add the remaining test cases to cover the whole function directly, which should be simpler then covering it through serializeToString.

Here is a failing test

	test('should jump over existing prefixes when generating new ones', () => {
		const doc = new DOMParser().parseFromString('<doc xmlns:ns1="uri:existing"/>', MIME_TYPE.XML_TEXT);
		const element = doc.documentElement;

		// Add an attribute that would normally use ns1 prefix
		element.setAttributeNS('uri:new', 'name', 'value');
		element.setAttributeNS('uri:other', 'ns1:other', 'value');

		const serialized = new XMLSerializer().serializeToString(doc);
		expect(serialized).toContain('xmlns:ns1="uri:existing"');
		expect(serialized).toContain('xmlns:ns2="uri:new"');
		expect(serialized).toContain('ns2:name="value"');
		expect(serialized).toContain('xmlns:ns3="uri:other"');
		expect(serialized).toContain('ns3:other="value"');
	});

The serialized result is:

<doc xmlns:ns1="uri:existing" xmlns:ns2="uri:new" ns2:name="value" xmlns:ns1="uri:other" ns1:other="value"/>

Notice the duplicate attribute.

I tried to work on the serializer but ended up completely rewriting it because I disliked the monolithic approach, and having it all squished into dom.js was quite cumbersome to work on.

This is what I ended up with

I had to fiddle with it quite a bit to match the expected behavior of the old serializer (attribute ordering, self closing tags etc) and managed to end up with only 1 failing test:

'should add local namespace as xmlns in HTML'
Expected: "<a:foo xmlns:a=\"AAA\"><child xmlns=\"AAA\"></child></a:foo>"
Received: "<a:foo xmlns:a=\"AAA\"><a:child></a:child></a:foo>"

But I would argue that what this test expects is wrong.

In any case I recognize this would be a big change so I am not going to push here or a new PR if this is not the direction you want to go in.