Allow attributes without namespace on jRuby

[flavorjones]

Discussed in #2675

^{Originally posted by ShockwaveNN October 20, 2022}
Hi, trying to port one of my library for parsing ooxml to jruby and I got a problem

I got this XML

<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<workbook xmlns="http://schemas.openxmlformats.org/spreadsheetml/2006/main"
          xmlns:r="http://schemas.openxmlformats.org/officeDocument/2006/relationships">
  <bookViews>
    <workbookView r:xWindow="360" yWindow="15" windowWidth="20955"
                  windowHeight="9720" activeTab="0"/>
  </bookViews>
</workbook>

And simplified ruby code to parse it:

require 'nokogiri'
xml = Nokogiri::XML(File.open('example.xml'), &:strict)
xml.xpath('xmlns:workbook/xmlns:bookViews/xmlns:workbookView').each do |sheet|
  puts sheet.attribute('xWindow').value
end

On mruby this output 360

But on jruby it just crash with:

NoMethodError: undefined method `value' for nil:NilClass

If I change puts to

  puts sheet.attribute('r:xWindow').value

It works correctly on both rubies

I understand that mruby and jruby uses different XML parser

Is there any option to jruby to make parsing of such attributes same as on jruby or proper way will be to fix my codebase by adding r: namespace?

[flavorjones]

Hi! I've reproduced what you're seeing, and just to share my analysis:

#! /usr/bin/env ruby

require "nokogiri"

xml = <<~XML
<?xml version="1.0" encoding="UTF-8" standalone="yes"?>
<workbook xmlns="http://schemas.openxmlformats.org/spreadsheetml/2006/main"
          xmlns:r="http://schemas.openxmlformats.org/officeDocument/2006/relationships">
  <bookViews>
    <workbookView r:xWindow="360" yWindow="15" windowWidth="20955"
                  windowHeight="9720" activeTab="0"/>
  </bookViews>
</workbook>
XML

doc = Nokogiri::XML(xml, &:strict)
sheet = doc.at_xpath('xmlns:workbook/xmlns:bookViews/xmlns:workbookView')
puts sheet.attribute('xWindow').inspect

On CRuby, this outputs something like:

#<Nokogiri::XML::Attr:0x50 name="xWindow" namespace=#<Nokogiri::XML::Namespace:0x3c prefix="r" href="http://schemas.openxmlformats.org/officeDocument/2006/relationships"> value="360">

but on JRuby, the attribute lookup fails and so the script prints nil.

This appears to be a bug, and I'm honestly surprised that we don't have tests for the desired behavior. I'll dig in!

[flavorjones]

OK, this isn't a simple fix, unfortunately, because the JRuby implementation is handling attributes inconsistently, sometimes treating local names as prefixed names, and vice versa, in a few different places.

I'll work on it, but it may not be fixed immediately.

[flavorjones]

See #2679 for the work-in-progress. Maybe this branch is good enough for you? If so, please let me know and I may be able to get that partial fix into the next release.