Consider unseen tracks in data source matching

beets/autotag/distance.py

@@ -409,9 +409,7 @@ def track_distance(
         dist.add_expr("medium", item.disc != track_info.medium)
 
     # Plugins.
-    if (original := item.get("data_source")) and (
-        actual := track_info.data_source
-    ) != original:
+    if (actual := track_info.data_source) != item.get("data_source"):
         dist.add("data_source", metadata_plugins.get_penalty(actual))
 
     return dist
@@ -529,9 +527,6 @@ def distance(
         dist.add("unmatched_tracks", 1.0)
 
     # Plugins.
-    if (
-        likelies["data_source"]
-        and (data_source := album_info.data_source) != likelies["data_source"]
-    ):
+    if (data_source := album_info.data_source) != likelies["data_source"]:
         dist.add("data_source", metadata_plugins.get_penalty(data_source))
     return dist

docs/plugins/index.rst

@@ -61,11 +61,6 @@ databases. They share the following configuration options:
   candidate comes from a different data source than what appears to be the
   original source of your music collection.
 
-  .. important::
-
-      This setting only applies to reimports, not to first-time imports, since
-      ``data_source`` is unknown for new files.
-
   **Example configurations:**
 
   .. code-block:: yaml

test/autotag/test_distance.py

@@ -330,7 +330,7 @@ class TestDataSourceDistance:
             _p("Original", "Other", 0.5, 1.0, MISMATCH, id="mismatch"),
             _p("Original", "unknown", 0.5, 1.0, MISMATCH, id="mismatch-unknown"),  # noqa: E501
             _p("Original", None, 0.5, 1.0, MISMATCH, id="mismatch-no-info"),
-            _p(None, "Other", 0.5, 1.0, MATCH, id="match-no-original"),
+            _p(None, "Other", 0.5, 1.0, MISMATCH, id="mismatch-no-original"),
             _p("unknown", "unknown", 0.5, 1.0, MATCH, id="match-unknown"),
             _p("Original", "Other", 1.0, 1.0, 0.25, id="mismatch-max-penalty"),
             _p("Original", "Other", 0.5, 5.0, 0.3125, id="mismatch-high-weight"),  # noqa: E501