implement smallest set rule application for titles

Author: CarsonDavis

sde_collections/models/delta_patterns.py

@@ -388,6 +388,35 @@ def generate_title_for_url(self, url_obj) -> tuple[str, str | None]:
         except (ValueError, ValidationError) as e:
             return None, str(e)
 
+    def get_url_match_count(self):
+        """
+        Get the number of unique URLs this pattern matches across both delta and curated URLs.
+        """
+        delta_urls = set(self.get_matching_delta_urls().values_list("url", flat=True))
+        curated_urls = set(self.get_matching_curated_urls().values_list("url", flat=True))
+        return len(delta_urls.union(curated_urls))
+
+    def is_most_distinctive_pattern(self, url) -> bool:
+        """
+        Determine if this pattern should apply to a URL by checking if it matches
+        the smallest number of URLs among all patterns that match this URL.
+        Returns True if this pattern should be applied.
+        """
+        my_match_count = self.get_url_match_count()
+
+        # Get all patterns that match this URL based on match_pattern regex
+        matching_patterns = DeltaTitlePattern.objects.filter(collection=self.collection).exclude(
+            id=self.id
+        )  # Exclude self to avoid duplicate counting
+
+        # Filter to only patterns that would match this URL and get their counts
+        for pattern in matching_patterns:
+            if re.match(pattern.get_regex_pattern(), url.url):
+                if pattern.get_url_match_count() < my_match_count:
+                    return False
+
+        return True
+
     def apply(self) -> None:
         """
         Apply the title pattern to matching URLs:
@@ -408,6 +437,9 @@ def apply(self) -> None:
 
         # Process each previously unaffected curated URL
         for curated_url in previously_unaffected_curated:
+            if not self.is_most_distinctive_pattern(curated_url):
+                continue
+
             new_title, error = self.generate_title_for_url(curated_url)
 
             if error:
@@ -439,15 +471,19 @@ def apply(self) -> None:
 
         # Update titles for all matching Delta URLs
         for delta_url in self.get_matching_delta_urls():
+            if not self.is_most_distinctive_pattern(delta_url):
+                continue
+
             new_title, error = self.generate_title_for_url(delta_url)
 
             if error:
                 DeltaResolvedTitleError.objects.create(title_pattern=self, delta_url=delta_url, error_string=error)
                 continue
 
-            # Update title and record resolution
+            # Update title and record resolution - key change here
             DeltaResolvedTitle.objects.update_or_create(
-                title_pattern=self, delta_url=delta_url, defaults={"resolved_title": new_title}
+                delta_url=delta_url,  # Only use delta_url for lookup
+                defaults={"title_pattern": self, "resolved_title": new_title},
             )
 
             delta_url.generated_title = new_title

sde_collections/tests/test_title_patterns.py

@@ -0,0 +1,79 @@
+# docker-compose -f local.yml run --rm django pytest sde_collections/tests/test_title_patterns.py
+
+import pytest
+
+from sde_collections.models.delta_patterns import DeltaResolvedTitle, DeltaTitlePattern
+from sde_collections.models.delta_url import DeltaUrl
+from sde_collections.tests.factories import CollectionFactory, DeltaUrlFactory
+
+
+@pytest.mark.django_db
+def test_title_pattern_multiple_resolved_titles_extended():
+    """Test that patterns properly handle title resolution based on URL set size."""
+    collection = CollectionFactory()
+
+    # Create URLs with different levels of specificity
+    url1 = DeltaUrlFactory(
+        collection=collection, url="https://example.com/docs/item.html", scraped_title="Original Title"
+    )
+    url2 = DeltaUrlFactory(
+        collection=collection, url="https://example.com/docs/item2.html", scraped_title="Original Title"
+    )
+    url3 = DeltaUrlFactory(
+        collection=collection, url="https://example.com/docs/pdfs/item1.html", scraped_title="Original Title"
+    )
+
+    # Create general pattern (matches all URLs)
+    general_pattern = DeltaTitlePattern.objects.create(
+        collection=collection,
+        match_pattern="*docs*",
+        title_pattern="{title} - Docs",
+        match_pattern_type=2,
+    )
+
+    # Verify initial pattern application
+    assert general_pattern.get_url_match_count() == 3
+    assert DeltaUrl.objects.get(pk=url1.pk).generated_title == "Original Title - Docs"
+    assert DeltaUrl.objects.get(pk=url2.pk).generated_title == "Original Title - Docs"
+    assert DeltaUrl.objects.get(pk=url3.pk).generated_title == "Original Title - Docs"
+
+    # Verify DeltaResolvedTitle entries
+    assert DeltaResolvedTitle.objects.count() == 3
+    for url in [url1, url2, url3]:
+        resolved = DeltaResolvedTitle.objects.get(delta_url=url)
+        assert resolved.title_pattern == general_pattern
+        assert resolved.resolved_title == "Original Title - Docs"
+
+    # Create more specific pattern
+    specific_pattern = DeltaTitlePattern.objects.create(
+        collection=collection, match_pattern="*docs/pdfs*", title_pattern="{title} - HTML", match_pattern_type=2
+    )
+
+    # Verify pattern match counts
+    assert specific_pattern.get_url_match_count() == 1  # Only matches pdfs URL
+    assert general_pattern.get_url_match_count() == 3  # Matches all URLs
+
+    # Verify titles were updated appropriately
+    assert DeltaUrl.objects.get(pk=url1.pk).generated_title == "Original Title - Docs"  # Unchanged
+    assert DeltaUrl.objects.get(pk=url2.pk).generated_title == "Original Title - Docs"  # Unchanged
+    assert DeltaUrl.objects.get(pk=url3.pk).generated_title == "Original Title - HTML"  # Updated
+
+    # Verify DeltaResolvedTitle entries
+    assert DeltaResolvedTitle.objects.count() == 3  # Still one per URL
+
+    # URLs with general pattern should be unchanged
+    for url in [url1, url2]:
+        resolved = DeltaResolvedTitle.objects.get(delta_url=url)
+        assert resolved.title_pattern == general_pattern
+        assert resolved.resolved_title == "Original Title - Docs"
+
+    # PDF URL should now use specific pattern
+    resolved_pdf = DeltaResolvedTitle.objects.get(delta_url=url3)
+    assert resolved_pdf.title_pattern == specific_pattern
+    assert resolved_pdf.resolved_title == "Original Title - HTML"
+
+    # Verify pattern relationships are maintained
+    assert url1 in general_pattern.delta_urls.all()
+    assert url2 in general_pattern.delta_urls.all()
+    assert url3 in general_pattern.delta_urls.all()
+    assert url3 in specific_pattern.delta_urls.all()