server: update similar post logic to yield more results

2 files changed, 39 insertions, 2 deletions

diff --git a/server/szurubooru/func/similar.py b/server/szurubooru/func/similar.py
index 77d7805..37f3f0d 100644
--- a/server/szurubooru/func/similar.py
+++ b/server/szurubooru/func/similar.py
@@ -14,7 +14,7 @@ def find_similar_posts(source_post: model.Post, limit: int) -> List[model.Post]:
     # This will help yield results quicker
     source_tags = sorted(source_post.tags, key=lambda t: t.post_count)
     source_tag_count = len(source_tags)
-    max_removals = ceil(source_tag_count / 2)  # remove at most 50% of tags
+    max_removals = source_tag_count - 1
 
     tags = source_tags
     for x in range(max_removals + 1):
diff --git a/server/szurubooru/tests/func/test_similar.py b/server/szurubooru/tests/func/test_similar.py
index 9200ff1..486f67b 100644
--- a/server/szurubooru/tests/func/test_similar.py
+++ b/server/szurubooru/tests/func/test_similar.py
@@ -9,9 +9,11 @@ def verify_posts():
         actual_post_ids = list([p.post_id for p in actual_posts])
         expected_post_ids = list([p.post_id for p in expected_posts])
         assert actual_post_ids == expected_post_ids
+
     return verify
 
 
+# I'd like my behavior to be like this
 def test_find_similar_posts(post_factory, tag_factory, verify_posts):
     tagA = tag_factory(names=["a"])
     tagB = tag_factory(names=["b"])
@@ -43,6 +45,38 @@ def test_find_similar_posts(post_factory, tag_factory, verify_posts):
     verify_posts(results, [postABC, postBC, postAB, postA])
 
 
+# but it's actually like this for performance reasons
+def test_find_similar_posts_naive(post_factory, tag_factory, verify_posts):
+    tagA = tag_factory(names=["a"])  # count=4
+    tagB = tag_factory(names=["b"])  # count=3
+    tagC = tag_factory(names=["c"])  # count=3
+    postA = post_factory(id=1, tags=[tagA])
+    postAB = post_factory(id=2, tags=[tagA, tagB])
+    postAC = post_factory(id=3, tags=[tagA, tagC])
+    postABC = post_factory(id=4, tags=[tagA, tagB, tagC])
+    postBC = post_factory(id=5, tags=[tagB, tagC])
+    db.session.add_all([tagA, tagB, tagC, postA, postAB, postAC, postABC, postBC])
+    db.session.flush()
+
+    results = similar.find_similar_posts(postBC, 10)
+    verify_posts(results, [postABC, postAC])
+
+    results = similar.find_similar_posts(postBC, 2)
+    verify_posts(results, [postABC, postAC])
+
+    results = similar.find_similar_posts(postABC, 10)
+    verify_posts(results, [postAC, postAB, postA])
+
+    results = similar.find_similar_posts(postA, 10)
+    verify_posts(results, [postABC, postAC, postAB])  # sorted by id
+
+    results = similar.find_similar_posts(postAB, 10)
+    verify_posts(results, [postABC, postAC, postA])
+
+    results = similar.find_similar_posts(postAC, 10)
+    verify_posts(results, [postABC, postAB, postA])
+
+
 def test_find_similar_posts_with_limit(post_factory, tag_factory, verify_posts):
     tagA = tag_factory(names=["a"])
     tagB = tag_factory(names=["b"])
@@ -56,4 +90,7 @@ def test_find_similar_posts_with_limit(post_factory, tag_factory, verify_posts):
     db.session.flush()
 
     results = similar.find_similar_posts(postABCDE, 10)
-    verify_posts(results, [postAB])
+    # I'd like it to be like this:
+    # verify_posts(results, [postAB])
+    # but it's like this for performance reasons:
+    verify_posts(results, [postAB, postA])