Merge pull request #2974 from nexB/2972-summary-consider-copyrights

Consider only copyrights in summry #2972

Author: pombredanne

src/summarycode/copyright_tallies.py

@@ -485,49 +485,50 @@ def filter_junk(texts):
 
 
 COMMON_NAMES = {
-    '3dfxinteractiveinc.': '3dfx Interactive, Inc.',
+    '3dfxinteractiveinc.': '3dfx Interactive',
 
     'cern': 'CERN - European Organization for Nuclear Research',
 
-    'ciscosystemsinc': 'Cisco Systems, Inc.',
-    'ciscosystems': 'Cisco Systems, Inc.',
-    'cisco': 'Cisco Systems, Inc.',
+    'ciscosystemsinc': 'Cisco Systems',
+    'ciscosystems': 'Cisco Systems',
+    'cisco': 'Cisco Systems',
 
-    'daisy': 'Daisy Ltd.',
+    'daisy': 'Daisy',
+    'daisyltd': 'Daisy',
 
-    'fsf': 'Free Software Foundation, Inc.',
-    'freesoftwarefoundation': 'Free Software Foundation, Inc.',
-    'freesoftwarefoundationinc': 'Free Software Foundation, Inc.',
-    'thefreesoftwarefoundation': 'Free Software Foundation, Inc.',
-    'thefreesoftwarefoundationinc': 'Free Software Foundation, Inc.',
+    'fsf': 'Free Software Foundation',
+    'freesoftwarefoundation': 'Free Software Foundation',
+    'freesoftwarefoundationinc': 'Free Software Foundation',
+    'thefreesoftwarefoundation': 'Free Software Foundation',
+    'thefreesoftwarefoundationinc': 'Free Software Foundation',
 
-    'hp': 'Hewlett-Packard, Inc.',
-    'hewlettpackard': 'Hewlett-Packard, Inc.',
-    'hewlettpackardco': 'Hewlett-Packard, Inc.',
-    'hpcompany': 'Hewlett-Packard, Inc.',
-    'hpdevelopmentcompanylp': 'Hewlett-Packard, Inc.',
-    'hpdevelopmentcompany': 'Hewlett-Packard, Inc.',
-    'hewlettpackardcompany': 'Hewlett-Packard, Inc.',
+    'hp': 'Hewlett-Packard',
+    'hewlettpackard': 'Hewlett-Packard',
+    'hewlettpackardco': 'Hewlett-Packard',
+    'hpcompany': 'Hewlett-Packard',
+    'hpdevelopmentcompanylp': 'Hewlett-Packard',
+    'hpdevelopmentcompany': 'Hewlett-Packard',
+    'hewlettpackardcompany': 'Hewlett-Packard',
 
-    'theandroidopensourceproject': 'The Android Open Source Project, Inc.',
-    'androidopensourceproject': 'The Android Open Source Project, Inc.',
+    'theandroidopensourceproject': 'Android Open Source Project',
+    'androidopensourceproject': 'Android Open Source Project',
 
-    'ibm': 'IBM Corporation',
+    'ibm': 'IBM',
 
-    'redhat': 'Red Hat, Inc.',
-    'redhatinc': 'Red Hat, Inc.',
+    'redhat': 'Red Hat',
+    'redhatinc': 'Red Hat',
 
-    'softwareinthepublicinterest': 'Software in the Public Interest, Inc.',
-    'spiinc': 'Software in the Public Interest, Inc.',
+    'softwareinthepublicinterest': 'Software in the Public Interest',
+    'spiinc': 'Software in the Public Interest',
 
-    'suse': 'SuSE, Inc.',
-    'suseinc': 'SuSE, Inc.',
+    'suse': 'SuSE',
+    'suseinc': 'SuSE',
 
-    'sunmicrosystems': 'Sun Microsystems, Inc.',
-    'sunmicrosystemsinc': 'Sun Microsystems, Inc.',
-    'sunmicro': 'Sun Microsystems, Inc.',
+    'sunmicrosystems': 'Sun Microsystems',
+    'sunmicrosystemsinc': 'Sun Microsystems',
+    'sunmicro': 'Sun Microsystems',
 
-    'thaiopensourcesoftwarecenter': 'Thai Open Source Software Center Ltd.',
+    'thaiopensourcesoftwarecenter': 'Thai Open Source Software Center',
 
     'apachefoundation': 'The Apache Software Foundation',
     'apachegroup': 'The Apache Software Foundation',
@@ -540,20 +541,19 @@ def filter_junk(texts):
 
     'regentsoftheuniversityofcalifornia': 'The Regents of the University of California',
 
-    # 'mit': 'the Massachusetts Institute of Technology',
-
-    'borland': 'Borland Corp.',
+    'borland': 'Borland',
+    'borlandcorp': 'Borland',
 
     'microsoft': 'Microsoft',
     'microsoftcorp': 'Microsoft',
     'microsoftinc': 'Microsoft',
     'microsoftcorporation': 'Microsoft',
 
-    'google': 'Google Inc.',
-    'googlellc': 'Google Inc.',
-    'googleinc': 'Google Inc.',
+    'google': 'Google',
+    'googlellc': 'Google',
+    'googleinc': 'Google',
 
-    'intel': 'Intel Corporation',
+    'intel': 'Intel',
 }
 
 # Remove everything except letters and numbers

src/summarycode/summarizer.py

@@ -18,6 +18,7 @@
 from cluecode.copyrights import CopyrightDetector
 from packagedcode.utils import combine_expressions
 from packagedcode import models
+from summarycode.copyright_tallies import canonical_holder
 from summarycode.score import compute_license_score
 from summarycode.score import get_field_values_from_codebase_resources
 from summarycode.score import unique
@@ -167,7 +168,7 @@ def get_declared_holders(codebase, holders_tallies):
         codebase, 'holders', key_files_only=True
     )
     entry_by_key_file_holders = {
-        fingerprints.generate(entry['holder']): entry
+        fingerprints.generate(canonical_holder(entry['holder'])): entry
         for entry in key_file_holders
         if entry['holder']
     }
@@ -212,25 +213,24 @@ def get_primary_language(programming_language_tallies):
 
 def get_origin_info_from_top_level_packages(top_level_packages, codebase):
     """
-    Return a 3-tuple containing the strings of declared license expression,
-    copyright holder, and primary programming language from a
+    Return a 3-tuple containing the declared license expression string, a list
+    of copyright holder, and primary programming language string from a
     ``top_level_packages`` list of detected top-level packages mapping and a
     ``codebase``.
     """
     if not top_level_packages:
-        return '', '', ''
+        return '', [], ''
 
     license_expressions = []
     programming_languages = []
     copyrights = []
-    parties = []
-
-    for package_mapping in top_level_packages:
-        package = models.Package.from_dict(package_mapping)
-        # we are only interested in key packages
-        if not is_key_package(package, codebase):
-            continue
 
+    top_level_packages = [
+        models.Package.from_dict(package_mapping)
+        for package_mapping in top_level_packages
+    ]
+    key_file_packages = [p for p in top_level_packages if is_key_package(p, codebase)]
+    for package in key_file_packages:
         license_expression = package.license_expression
         if license_expression:
             license_expressions.append(license_expression)
@@ -243,8 +243,6 @@ def get_origin_info_from_top_level_packages(top_level_packages, codebase):
         if copyright_statement:
             copyrights.append(copyright_statement)
 
-        parties.extend(package.parties or [])
-
     # Combine license expressions
     unique_license_expressions = unique(license_expressions)
     combined_declared_license_expression = combine_expressions(
@@ -263,9 +261,20 @@ def get_origin_info_from_top_level_packages(top_level_packages, codebase):
     declared_holders = []
     if holders:
         declared_holders = holders
-    elif parties:
-        declared_holders = [party.name for party in parties or []]
-
+    else:
+        # If the package data does not contain an explicit copyright, check the
+        # key files where the package data was detected from and see if there
+        # are any holder detections that can be used.
+        for package in key_file_packages:
+            for datafile_path in package.datafile_paths:
+                key_file_resource = codebase.get_resource(path=datafile_path)
+                if not key_file_resource:
+                    continue
+                holders = [h['holder'] for h in key_file_resource.holders]
+                declared_holders.extend(holders)
+    # Normalize holder names before collecting them
+    # This allows us to properly remove declared holders from `other_holders` later
+    declared_holders = [canonical_holder(h) for h in declared_holders]
     declared_holders = unique(declared_holders)
 
     # Programming language

tests/cluecode/data/copyrights/COPYING_gpl-COPYING_gpl.gpl.yml

@@ -9,5 +9,5 @@ holders:
   - Free Software Foundation, Inc.
   - the Free Software Foundation
 holders_summary:
-  - value: Free Software Foundation, Inc.
+  - value: Free Software Foundation
     count: 2

tests/cluecode/data/copyrights/afferogplv1-AfferoGPLv.yml

@@ -13,5 +13,5 @@ holders:
 holders_summary:
   - value: Affero Inc.
     count: 2
-  - value: Free Software Foundation, Inc.
+  - value: Free Software Foundation
     count: 1

tests/cluecode/data/copyrights/afferogplv3-AfferoGPLv.yml

@@ -7,5 +7,5 @@ copyrights:
 holders:
   - Free Software Foundation, Inc.
 holders_summary:
-  - value: Free Software Foundation, Inc.
+  - value: Free Software Foundation
     count: 1

tests/cluecode/data/copyrights/android_c-c.c.yml

@@ -9,7 +9,7 @@ holders:
   - The Android Open Source Project
   - Colin Percival
 holders_summary:
-  - value: Colin Percival
+  - value: Android Open Source Project
     count: 1
-  - value: The Android Open Source Project, Inc.
+  - value: Colin Percival
     count: 1

tests/cluecode/data/copyrights/apache2_debian_trailing_name_missed-apache.label.yml

@@ -75,7 +75,7 @@ holders_summary:
     count: 2
   - value: Board of Trustees of the University of Illinois
     count: 1
-  - value: Cisco Systems, Inc.
+  - value: Cisco Systems
     count: 1
   - value: Eric Haines
     count: 1

tests/cluecode/data/copyrights/bigelow_holmes-Bigelow&Holmes.yml

@@ -11,5 +11,5 @@ holders:
 holders_summary:
   - value: Bigelow & Holmes
     count: 1
-  - value: Sun Microsystems, Inc.
+  - value: Sun Microsystems
     count: 1

tests/cluecode/data/copyrights/colin_android-bsdiff_c.c.yml

@@ -9,7 +9,7 @@ holders:
   - The Android Open Source Project
   - Colin Percival
 holders_summary:
-  - value: Colin Percival
+  - value: Android Open Source Project
     count: 1
-  - value: The Android Open Source Project, Inc.
+  - value: Colin Percival
     count: 1

tests/cluecode/data/copyrights/complex_notice-NOTICE.yml

@@ -58,7 +58,7 @@ holders:
 holders_summary:
   - value: David Schultz
     count: 9
-  - value: Sun Microsystems, Inc.
+  - value: Sun Microsystems
     count: 4
   - value: Mike Barcroft
     count: 2