Fix overlap percentage calculation (#1668)

* Fix overlap calculation

* Add test for initialization of section_overlap

Author: bastbu

app/backend/prepdocslib/textsplitter.py

@@ -93,7 +93,7 @@ def __init__(self, has_image_embeddings: bool, max_tokens_per_section: int = 500
         self.max_section_length = DEFAULT_SECTION_LENGTH
         self.sentence_search_limit = 100
         self.max_tokens_per_section = max_tokens_per_section
-        self.section_overlap = self.max_section_length // DEFAULT_OVERLAP_PERCENT
+        self.section_overlap = int(self.max_section_length * DEFAULT_OVERLAP_PERCENT / 100)
         self.has_image_embeddings = has_image_embeddings
 
     def split_page_by_max_tokens(self, page_num: int, text: str) -> Generator[SplitPage, None, None]:

tests/test_sentencetextsplitter.py

@@ -0,0 +1,23 @@
+from unittest.mock import patch
+
+import pytest
+
+from prepdocslib.textsplitter import SentenceTextSplitter
+
+
+@pytest.mark.parametrize(
+    "actual_percentage, expected_section_overlap",
+    [
+        (100, 1000),
+        (80, 800),
+        (10.75, 107),
+        (10, 100),
+        (0, 0),
+    ],
+)
+def test_sentence_text_splitter_initializes_overlap_correctly(
+    actual_percentage: float, expected_section_overlap: float
+):
+    with patch("prepdocslib.textsplitter.DEFAULT_OVERLAP_PERCENT", actual_percentage):
+        subject = SentenceTextSplitter(False)
+        assert subject.section_overlap == expected_section_overlap