Merge pull request #65 from climatepolicyradar/RND-490-text-block-has…

…hable

RND-490: making TextBlock objects hashable

src/cpr_data_access/models.py

@@ -196,14 +196,19 @@ def to_string(self) -> str:
         """Return text in a clean format"""
         return " ".join([line.strip() for line in self.text])
 
+    def __hash__(self) -> int:
+        """Get hash of the text-block. Based on the text and the text_block_id"""
+        text_utf8 = self.to_string().encode("utf-8")
+
+        return hash(f"{text_utf8}-{self.text_block_id.encode()}")
+
     @cached_property
     def text_hash(self) -> str:
         """
         Get hash of text block text. If the text block has no text (although this shouldn't be the case), return an empty string.
 
         :return str: md5sum + "__" + sha256, or empty string if the text block has no text
         """
-
         if self.text == "":
             return ""
 

tests/test_models.py

@@ -13,6 +13,7 @@
     CPRDocumentMetadata,
     Span,
     KnowledgeBaseIDs,
+    TextBlock,
 )
 
 
@@ -433,3 +434,20 @@ def test_display_text_block(test_document, test_spans_valid):
     assert isinstance(block_html, str)
     assert len(block_html) > 0
     assert block_html.startswith("<div")
+
+
+def test_text_block_hashable(test_document):
+    doc = test_document
+
+    set(doc.text_blocks)
+
+    first_block_hash = doc.text_blocks[0].__hash__()
+    assert isinstance(first_block_hash, int)
+
+    comparison_block = TextBlock(**doc.text_blocks[0].dict())
+
+    assert comparison_block == doc.text_blocks[0]
+
+    comparison_block.text_block_id = "0"
+
+    assert comparison_block != doc.text_blocks[0]