Simplify element.attributes structure

[tompng]

Basically refactor Attributes, and fixes some bugs found while doing it.

Structure of Attributes, a subclass of Hash

For XML: <element name1="1" ns1:name2="2" ns2:name2="3" ns1:name3="4"/>, the structure was:

{
  name1 => attr1
  name2 => { ns1 => attr2, ns2 => attr3 }
  name3 => att4
}

This structure is useful for fuzzy name search, and fuzzy name deletion, but we don't need this complicated structure because fuzzy search should be removed, and fuzzy deletion is not used from REXML (and it wasn't working at all)

I think the mental model of namespaced attribute in Document Object Model is something like this.

{ default: { name1: attr1 }, ns1: { name2: attr1, name3: attr4}, ns2: { name2: attr2 } }
# or simply
{ "name1": attr1, "ns1:name2": attr2, "ns2: name2": attr3, "ns1:name3": attr4 }

This PR changes it to the latter, simple one.

Stop fuzzy get_attribute/get_attribute_ns search

Attributes#get_attribute does fuzzy search only on its own attribute, not on attributes defined in ATTLIST. Attributes#get_attribute_ns tries fuzzy search but fails because of a bug.
In Document Object Model, get_attribute and get_attribute_ns should perform strictly no-prefix attribute if prefix/namespace is not given.

ATTLIST lookup

There's many redundant duplicated code for ATTLIST lookup.
Attributes#get_attribute_ns forgot to check ATTLIST (bug).
Introduce each_effective_attribute that iterates attributes defined in ATTLIST and element's own attributes

Fixed bugs

• get_attribute: Fuzzy search shouldn't be performed
• get_attribute_ns: forgot to lookup ATTLIST
• delete_all: Unlike the name that indicates fuzzy deletion, it deletes only one attribute, and returns wrong attrribute which is a different one than actually deleted when prefix is given.
• namespace declaration check: foo:xmnls was wrongly considered namespace declaration
• And perhaps more which I haven't found yet

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

This PR refactors REXML attribute handling to simplify lookup/storage while reintroducing ATTLIST (DTD) default attribute support, and updates/extends tests to match the new behavior.

Changes:

• Simplify attribute storage/access in REXML::Element::Attributes, adding support utilities for ATTLIST defaults and namespace-declaration detection.
• Add Doctype#attribute_declarations_of and refactor existing DTD attribute accessors to use it.
• Update and add tests around attribute deletion, namespace-qualified lookups, and attributes.to_h expectations.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
test/test_core.rb	Updates expectations for attribute hashing and adds backward-compat coverage for DTD attribute accessors.
test/test_contrib.rb	Adjusts tests to use fully qualified attribute names for namespaced attributes.
test/test_attributes.rb	Adds coverage for delete_all behavior across namespaces.
lib/rexml/element.rb	Refactors attribute iteration/lookup/deletion and introduces “effective attributes” merging element + DTD defaults.
lib/rexml/doctype.rb	Adds attribute_declarations_of and refactors attributes_of / attribute_of to use it.
lib/rexml/attribute.rb	Changes removal semantics and adds namespace_declaration? helper.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated 5 comments.

[Copilot]

Pull request overview

Copilot reviewed 8 out of 8 changed files in this pull request and generated 3 comments.

[naitoh]

I think this PR is reasonable corrections.
Thank you!