Add from-file attribute to matches with origin

Adds a new attribute to document the origin path from matches
so we can determine in the cases of de-referenced matches, the
location these came from, and also to differentiate between
matches that come from the same file or from a different file.

Signed-off-by: Ayan Sinha Mahapatra <[email protected]>

src/licensedcode/detection.py

@@ -617,6 +617,7 @@ def from_dict(cls, license_match_mapping):
         matched_text = license_match_mapping.get("matched_text") or None
 
         return cls(
+            from_file=license_match_mapping["from_file"],
             start_line=license_match_mapping["start_line"],
             end_line=license_match_mapping["end_line"],
             match_score=license_match_mapping["score"],
@@ -654,6 +655,7 @@ def to_dict(
 
         # Detection Level Information
         result['score'] = self.score()
+        result['from_file'] = self.from_file
         result['start_line'] = self.start_line
         result['end_line'] = self.end_line
         result['matched_length'] = self.len()
@@ -671,6 +673,21 @@ def to_dict(
         return result
 
 
+def populate_matches_with_path(matches, path):
+    """
+    Given `matches` list of LicenseMatch objects, populate the `from_file`
+    attribute in them with `path` which is the path for the origin file for
+    that license match.
+    """
+    for match in matches:
+        # Here if we have the `from_file` attribute populated already,
+        # they are from other files, and if it's empty, they are from
+        # the original resource, so we populate the files with the resource
+        # path for the original resource of their origin  
+        if not match["from_file"]:
+            match["from_file"] = path
+
+
 def collect_license_detections(codebase, include_license_clues=True):
     """
     Return a list of LicenseDetectionFromResult object rehydrated from
@@ -680,7 +697,10 @@ def collect_license_detections(codebase, include_license_clues=True):
     according to their license detections. This is required because package fields
     are populated in package plugin, which runs before the license plugin, and thus
     the license plugin step where unknown references to other files are dereferenced
-    does not show up automatically in package attributes. 
+    does not show up automatically in package attributes.
+
+    Also populate from_file attributes with resource paths for matches which have
+    origin in the same file.
     """
     has_packages = hasattr(codebase.root, 'package_data')
     has_licenses = hasattr(codebase.root, 'license_detections')
@@ -692,7 +712,11 @@ def collect_license_detections(codebase, include_license_clues=True):
         resource_license_detections = []
         if has_licenses:
             license_detections = getattr(resource, 'license_detections', []) or []
+            for detection in license_detections:
+                populate_matches_with_path(matches=detection["matches"], path=resource.path)
             license_clues = getattr(resource, 'license_clues', []) or []
+            populate_matches_with_path(matches=license_clues, path=resource.path)
+            codebase.save_resource(resource)
 
             if license_detections:
                 license_detection_objects = detections_from_license_detection_mappings(
@@ -729,6 +753,9 @@ def collect_license_detections(codebase, include_license_clues=True):
 
                 package_license_detections = package["license_detections"]
                 if package_license_detections:
+                    for detection in package_license_detections:
+                        populate_matches_with_path(matches=detection["matches"], path=resource.path)
+                        modified = True
                     package_license_detection_mappings.extend(package_license_detections)
                     detection_is_same, license_expression = verify_package_license_expression(
                         license_detection_mappings=package_license_detections,

src/licensedcode/match.py

@@ -223,6 +223,17 @@ class LicenseMatch(object):
         metadata=dict(help='match end line, 1-based')
     )
 
+    from_file = attr.ib(
+        default=None,
+        metadata=dict(
+            help='File path where this LicenseMatch was originally detected. '
+                 'This needs to be stored as we bring over LicenseMatches from '
+                 'other files into LicenseDetection objects now, and we need '
+                 'to track the origin for these to be able to determine easily '
+                 'which are native to that file.'
+        )
+    )
+
     query = attr.ib(
         default=None,
         metadata=dict(help='Query object for this match')
@@ -722,7 +733,7 @@ def matched_text(
         highlight=True,
         highlight_matched='{}',
         highlight_not_matched='[{}]',
-        _usecache=True
+        _usecache=True,
     ):
         """
         Return the matched text for this match or an empty string if no query
@@ -763,6 +774,7 @@ def to_dict(
         include_text=False,
         license_text_diagnostics=False,
         whole_lines=True,
+        file_path=None,
     ):
         """
         Return a "result" scan data built from a LicenseMatch object.
@@ -783,6 +795,7 @@ def to_dict(
         result['score'] = self.score()
         result['start_line'] = self.start_line
         result['end_line'] = self.end_line
+        result['from_file'] = file_path
         result['matched_length'] = self.len()
         result['match_coverage'] = self.coverage()
         result['matcher'] = self.matcher

src/licensedcode/plugin_license.py

@@ -20,6 +20,7 @@
 
 from licensedcode.cache import build_spdx_license_expression, get_cache
 from licensedcode.detection import collect_license_detections
+from licensedcode.detection import populate_matches_with_path
 from licensedcode.detection import find_referenced_resource
 from licensedcode.detection import get_detected_license_expression
 from licensedcode.detection import get_matches_from_detection_mappings
@@ -279,11 +280,14 @@ def add_referenced_filenames_license_matches_for_detections(resource, codebase):
                 modified = True
                 detection_modified = True
                 detections_added.extend(referenced_resource.license_detections)
-                license_match_mappings.extend(
-                    get_matches_from_detection_mappings(
-                        license_detections=referenced_resource.license_detections
-                    )
+                matches_to_extend = get_matches_from_detection_mappings(
+                    license_detections=referenced_resource.license_detections
                 )
+                populate_matches_with_path(
+                    matches=matches_to_extend,
+                    path=referenced_resource.path
+                )
+                license_match_mappings.extend(matches_to_extend)
 
         if not detection_modified:
             continue

src/packagedcode/licensing.py

@@ -25,6 +25,7 @@
 from licensedcode.detection import find_referenced_resource
 from licensedcode.detection import detect_licenses
 from licensedcode.detection import LicenseDetectionFromResult
+from licensedcode.detection import populate_matches_with_path
 from licensedcode.spans import Span
 from licensedcode import query
 
@@ -113,11 +114,16 @@ def add_referenced_license_matches_for_package(resource, codebase):
                 if referenced_license_detections:
                     modified = True
                     detection_modified = True
-                    license_match_mappings.extend(
-                        get_matches_from_detection_mappings(
-                            license_detections=referenced_license_detections
-                        )
+                    matches_to_extend = get_matches_from_detection_mappings(
+                        license_detections=referenced_license_detections
                     )
+                    # For LicenseMatches with different resources as origin, add the
+                    # resource path to these matches as origin info
+                    populate_matches_with_path(
+                        matches=matches_to_extend,
+                        path=referenced_resource.path
+                    )
+                    license_match_mappings.extend(matches_to_extend)
 
             if not detection_modified:
                 continue
@@ -231,6 +237,10 @@ def add_referenced_license_detection_from_package(resource, codebase):
                 for pkg_detection in pkg_detections:
                     modified = True
                     detection_modified = True
+                    populate_matches_with_path(
+                        matches=pkg_detection["matches"],
+                        path=resource.path
+                    )
                     license_match_mappings.extend(pkg_detection["matches"])
                     detections_added.append(pkg_detection)
                     analysis = DetectionCategory.UNKNOWN_REFERENCE_IN_FILE_TO_PACKAGE.value
@@ -347,6 +357,11 @@ def get_license_detections_from_sibling_file(resource, codebase):
 
     license_detections = []
     for sibling in siblings:
+        for detection in sibling.license_detections:
+            populate_matches_with_path(
+                matches=detection["matches"],
+                path=sibling.path
+            )
         license_detections.extend(sibling.license_detections)
 
     if not license_detections:

tests/cluecode/data/plugin_filter_clues/filtered-expected.json

@@ -36,6 +36,7 @@
               "score": 96.07,
               "start_line": 7,
               "end_line": 70,
+              "from_file": "LICENSE",
               "matched_length": 367,
               "match_coverage": 100.0,
               "matcher": "3-seq",

tests/cluecode/data/plugin_filter_clues/filtered-expected2.json

@@ -36,6 +36,7 @@
               "score": 100.0,
               "start_line": 7,
               "end_line": 22,
+              "from_file": "LICENSE2",
               "matched_length": 145,
               "match_coverage": 100.0,
               "matcher": "2-aho",

tests/cluecode/data/plugin_filter_clues/filtered-expected3.json

@@ -36,6 +36,7 @@
               "score": 100.0,
               "start_line": 1,
               "end_line": 47,
+              "from_file": "LICENSE3",
               "matched_length": 303,
               "match_coverage": 100.0,
               "matcher": "1-hash",

tests/formattedcode/data/common/manifests-expected.json

@@ -34,6 +34,7 @@
               "score": 100.0,
               "start_line": 1,
               "end_line": 1,
+              "from_file": null,
               "matched_length": 8,
               "match_coverage": 100.0,
               "matcher": "2-aho",
@@ -47,6 +48,7 @@
               "score": 100.0,
               "start_line": 2,
               "end_line": 2,
+              "from_file": null,
               "matched_length": 7,
               "match_coverage": 100.0,
               "matcher": "2-aho",
@@ -125,6 +127,7 @@
               "score": 100.0,
               "start_line": 1,
               "end_line": 1,
+              "from_file": null,
               "matched_length": 3,
               "match_coverage": 100.0,
               "matcher": "1-hash",
@@ -144,6 +147,7 @@
               "score": 100.0,
               "start_line": 1,
               "end_line": 1,
+              "from_file": null,
               "matched_length": 3,
               "match_coverage": 100.0,
               "matcher": "1-hash",
@@ -218,6 +222,7 @@
               "score": 100.0,
               "start_line": 1,
               "end_line": 1,
+              "from_file": null,
               "matched_length": 1,
               "match_coverage": 100.0,
               "matcher": "1-spdx-id",
@@ -640,6 +645,7 @@
                   "score": 100.0,
                   "start_line": 1,
                   "end_line": 1,
+                  "from_file": "manifests/maven/persistence-api-1.0.pom",
                   "matched_length": 8,
                   "match_coverage": 100.0,
                   "matcher": "2-aho",
@@ -653,6 +659,7 @@
                   "score": 100.0,
                   "start_line": 2,
                   "end_line": 2,
+                  "from_file": "manifests/maven/persistence-api-1.0.pom",
                   "matched_length": 7,
                   "match_coverage": 100.0,
                   "matcher": "2-aho",
@@ -697,6 +704,7 @@
               "score": 16.0,
               "start_line": 17,
               "end_line": 19,
+              "from_file": "manifests/maven/persistence-api-1.0.pom",
               "matched_length": 3,
               "match_coverage": 100.0,
               "matcher": "2-aho",
@@ -709,6 +717,7 @@
               "score": 82.35,
               "start_line": 18,
               "end_line": 20,
+              "from_file": "manifests/maven/persistence-api-1.0.pom",
               "matched_length": 14,
               "match_coverage": 82.35,
               "matcher": "3-seq",
@@ -829,6 +838,7 @@
                   "score": 100.0,
                   "start_line": 1,
                   "end_line": 1,
+                  "from_file": "manifests/npm-license-mapping/package.json",
                   "matched_length": 3,
                   "match_coverage": 100.0,
                   "matcher": "1-hash",
@@ -848,6 +858,7 @@
                   "score": 100.0,
                   "start_line": 1,
                   "end_line": 1,
+                  "from_file": "manifests/npm-license-mapping/package.json",
                   "matched_length": 3,
                   "match_coverage": 100.0,
                   "matcher": "1-hash",
@@ -991,6 +1002,7 @@
               "score": 100.0,
               "start_line": 6,
               "end_line": 6,
+              "from_file": "manifests/npm-license-mapping/package.json",
               "matched_length": 4,
               "match_coverage": 100.0,
               "matcher": "2-aho",
@@ -1009,6 +1021,7 @@
               "score": 100.0,
               "start_line": 20,
               "end_line": 20,
+              "from_file": "manifests/npm-license-mapping/package.json",
               "matched_length": 3,
               "match_coverage": 100.0,
               "matcher": "2-aho",
@@ -1133,6 +1146,7 @@
                   "score": 100.0,
                   "start_line": 1,
                   "end_line": 1,
+                  "from_file": "manifests/npm-license-string/package.json",
                   "matched_length": 1,
                   "match_coverage": 100.0,
                   "matcher": "1-spdx-id",
@@ -1226,6 +1240,7 @@
               "score": 100.0,
               "start_line": 4,
               "end_line": 4,
+              "from_file": "manifests/npm-license-string/package.json",
               "matched_length": 2,
               "match_coverage": 100.0,
               "matcher": "2-aho",
@@ -1354,6 +1369,7 @@
                   "score": 100.0,
                   "start_line": 1,
                   "end_line": 1,
+                  "from_file": "manifests/pypi/bluepyopt_setup.py",
                   "matched_length": 1,
                   "match_coverage": 100.0,
                   "matcher": "1-hash",
@@ -1373,6 +1389,7 @@
                   "score": 100.0,
                   "start_line": 1,
                   "end_line": 1,
+                  "from_file": "manifests/pypi/bluepyopt_setup.py",
                   "matched_length": 10,
                   "match_coverage": 100.0,
                   "matcher": "1-hash",
@@ -1524,6 +1541,7 @@
               "score": 100.0,
               "start_line": 9,
               "end_line": 20,
+              "from_file": "manifests/pypi/bluepyopt_setup.py",
               "matched_length": 106,
               "match_coverage": 100.0,
               "matcher": "2-aho",
@@ -1542,6 +1560,7 @@
               "score": 100.0,
               "start_line": 65,
               "end_line": 65,
+              "from_file": "manifests/pypi/bluepyopt_setup.py",
               "matched_length": 2,
               "match_coverage": 100.0,
               "matcher": "2-aho",
@@ -1560,6 +1579,7 @@
               "score": 100.0,
               "start_line": 74,
               "end_line": 75,
+              "from_file": "manifests/pypi/bluepyopt_setup.py",
               "matched_length": 10,
               "match_coverage": 100.0,
               "matcher": "2-aho",