Normalize licenses parsed from RO-Crate to SPDX ID

Author: fbacall

app/validators/license_validator.rb

@@ -3,12 +3,10 @@ class LicenseValidator < ActiveModel::EachValidator
   def validate_each(record, attribute, value)
     return if Seek::License.find(value)
     # Try looking up by URI
-    if value.start_with? /https?:/
-      id = Seek::License.uri_to_id(value)
-      if id
-        record.send("#{attribute}=", id)
-        return
-      end
+    license = Seek::License.normalize(license)
+    if license
+      record.send("#{attribute}=", license)
+      return
     end
     record.errors.add(attribute, options[:message] || "isn't a recognized license")
   end

lib/seek/license.rb

@@ -41,13 +41,24 @@ def self.find(id, source = Seek::License.combined)
     end
 
     def self.uri_to_id(uri)
-      uri_map[uri]
+      uri_map[uri.downcase]
     end
 
     def self.find_as_hash(id, source = Seek::License.combined)
       source[id]
     end
 
+    def self.normalize(license)
+      return nil if license.blank?
+      license = license.strip
+      if license.start_with?(/https?:/)
+        license = uri_to_id(license)
+      else
+        license = normalized_id_map[license.downcase]
+      end
+      license
+    end
+
     def is_null_license?
       id == NULL_LICENSE
     end
@@ -84,10 +95,19 @@ def self.uri_map
       @uri_map = {}
       combined.each do |id, license|
         (license['urls'] || []).each do |url|
-          @uri_map[url] ||= id
+          @uri_map[url.downcase] ||= id
         end
       end
       @uri_map
     end
+
+    def self.normalized_id_map
+      return @normalized_id_map if @normalized_id_map
+      @normalized_id_map = {}
+      combined.each_key do |id|
+        @normalized_id_map[id.downcase] = id
+      end
+      @normalized_id_map
+    end
   end
 end

lib/seek/workflow_extractors/cff.rb

@@ -39,7 +39,7 @@ def metadata
         metadata[:other_creators] = other_creators.join(', ')
 
         metadata[:title] = cff.title if cff.title.present?
-        metadata[:license] = cff.license if cff.license.present?
+        metadata[:license] = Seek::License.normalize(cff.license) if cff.license.present?
         metadata[:tags] = cff.keywords.map(&:strip) if cff.keywords.present?
         metadata[:doi] = cff.doi if cff.doi.present?
         metadata[:source_link_url] = cff.url if cff.url.present?

lib/seek/workflow_extractors/cwl.rb

@@ -100,7 +100,7 @@ def parse_metadata(existing_metadata, yaml_or_json_string)
           existing_metadata[:description] = cwl['doc']
         end
         if cwl.key?('s:license')
-          existing_metadata[:license] = cwl['s:license']
+          existing_metadata[:license] = Seek::License.normalize(cwl['s:license'])
         end
 
         existing_metadata[:internals] = {

lib/seek/workflow_extractors/galaxy.rb

@@ -39,7 +39,7 @@ def metadata
         end
 
         metadata[:description] = galaxy['annotation'] if galaxy['annotation'].present?
-        metadata[:license] = galaxy['license'] if galaxy['license'].present?
+        metadata[:license] = Seek::License.normalize(galaxy['license']) if galaxy['license'].present?
 
         if galaxy['creator']
           people, others = Array(galaxy['creator']).partition { |c| c['class'] == 'Person' }

lib/seek/workflow_extractors/rocrate.rb

@@ -74,7 +74,7 @@ def metadata_from_crate(crate, m)
 
         m[:title] = crate['name'] if crate['name'].present?
         m[:description] = crate['description'] if crate['description'].present?
-        m[:license] = crate['license'] if crate['license'].present?
+        m[:license] = Seek::License.normalize(crate['license']) if crate['license'].present?
 
         other_creators = []
         authors = []

test/fixtures/files/workflows/ro-crate-with-other-license-file.crate.zip

@@ -171,4 +171,18 @@ class LicenseTest < ActiveSupport::TestCase
     assert_equal 'No license - no permission to use unless the owner grants a licence', license.full_display_title
   end
 
+  test 'normalize incoming license' do
+    assert_equal 'CC-BY-4.0', Seek::License.normalize('https://spdx.org/licenses/CC-BY-4.0.html')
+    assert_equal 'CC-BY-4.0', Seek::License.normalize('https://spdx.org/licenses/CC-BY-4.0')
+    assert_equal 'CC-BY-4.0', Seek::License.normalize('https://creativecommons.org/licenses/by/4.0/')
+    assert_equal 'CC-BY-4.0', Seek::License.normalize('https://creativecommons.org/licenses/by/4.0/legalcode')
+    assert_equal 'CC-BY-4.0', Seek::License.normalize('CC-BY-4.0')
+    assert_equal 'CC-BY-4.0', Seek::License.normalize('cc-by-4.0')
+    assert_equal 'CC-BY-4.0', Seek::License.normalize('cc-By-4.0')
+    assert_equal 'CC-BY-4.0', Seek::License.normalize("         cc-By-4.0\n  ")
+    assert_nil Seek::License.normalize('huh what')
+    assert_nil Seek::License.normalize(nil)
+    assert_nil Seek::License.normalize('')
+  end
+
 end

test/unit/license_test.rb

@@ -127,6 +127,20 @@ class RoCrateExtractionTest < ActiveSupport::TestCase
     assert_equal 'Apache-2.0', metadata[:license]
   end
 
+  test 'extracts and normalizes URI license from ro-crate-metadata, and ignores Licensee "other" license file' do
+    # Checks 2 things:
+    # 1. A URI license, e.g. https://opensource.org/licenses/MIT, is normalized to its SPDX ID
+    #   (which means the correct license will be selected in the dropdown).
+    # 2. The SPDX ID `NOASSERTION`, returned by Licensee (which flags the file `not_a_license.py` as potentially containing a license),
+    #    is ignored in favour of the MIT license from the RO-Crate metadata.
+
+    wf = open_fixture_file('workflows/ro-crate-with-other-license-file.crate.zip')
+    extractor = Seek::WorkflowExtractors::ROCrate.new(wf)
+    metadata = extractor.metadata
+
+    assert_equal 'MIT', metadata[:license]
+  end
+
   test 'extracts author with affiliation as array' do
     # Also note that the following is just a regular RO-Crate, not a Workflow RO-Crate
     wf = open_fixture_file('ro_crates/affiliation_array.crate.zip')