XML zip imports look for extracted files in the wrong directory

[laritakr]

Summary

When an XmlParser import is given a zip file (parser_fields['import_file_path'] pointing at a .zip), the job extracts the zip into importer_unzip_path but XmlParser#file_paths and #metadata_paths then glob in the wrong directory, so the parser can't find the extracted XML files and the import fails (or finds no records).

• March 2020, PR Generic XML Importer #142: XmlParser introduced with the flawed File.dirname glob. Zip uploads glob the upload staging dir. Broken from day one.
• September 2021, PR Fix file paths for Zip CSV imports  #261: Base class real_import_file_path gains the importer_unzip_path if file? && zip? branch. XmlParser zip uploads now glob the parent of the unzip dir. Different broken location.

Trace

Post-extraction state:

• parser_fields['import_file_path'] = /tmp/imports/.../bundle.zip (unchanged from upload).
• importer_unzip_path = /tmp/imports/.../import_{path_string}/ (contains the extracted content).
• ApplicationParser#real_import_file_path returns importer_unzip_path when file? && zip?.
• XmlParser#import_file_path memoizes that directory path.

Then at app/parsers/bulkrax/xml_parser.rb lines 82-90:

def file_paths
  @file_paths ||=
    if file?
      Dir.glob("#{File.dirname(import_file_path)}/**/*").reject { |f| File.file?(f) == false }
    else
      Dir.glob("#{import_file_path}/**/*").reject { |f| File.file?(f) == false }
    end
end

• file? (in app/models/concerns/bulkrax/importer_exporter_behavior.rb:47) checks parser_fields['import_file_path'] — the raw zip path, which is a file → true.
• Line 85 takes the if branch → File.dirname(import_file_path) = File.dirname(importer_unzip_path) = the parent of the unzip dir.
• Glob runs against the parent, not the unzip dir itself — misses the extracted XML files.

Reproduction

Upload a zip containing metadata.xml and data files via the legacy importers UI with Parser = Bulkrax::XmlParser. The import will run but find zero metadata records or files.

Why this hasn't been caught

No specs cover XmlParser with a zip upload (spec/parsers/bulkrax/xml_parser_spec.rb doesn't exercise zip paths). No zip fixtures for XML exist. The locale hint at config/locales/bulkrax.en.yml:395 advertises XML zip support, but it doesn't actually work.

Proposed fix directions

Two options, either is viable:

1. Update ImporterJob#unzip_imported_file to set parser_fields['import_file_path'] to the extracted location after unzipping, so file?/zip? checks see the post-extraction state. More invasive; touches the core flow.

2. Override XmlParser#file_paths / #metadata_paths (or #real_import_file_path) to resolve relative to importer_unzip_path directly when a zip was uploaded. Localized to XmlParser.

Pre-existing bug (at least since PR #261 in Sep 2021); the current unzip-fix branch preserves the status quo rather than fixing it in scope.

Acceptance criteria

• Uploading a zip containing XML metadata results in the XML files being found and imported.
• Spec coverage: spec/parsers/bulkrax/xml_parser_spec.rb exercises the zip-upload flow end-to-end with a real fixture zip.
• Existing non-zip XML import behavior unchanged.