⚡ Fix quadratic Library.add during parsing and slow quote-enclosing check

bibtexparser/library.py

@@ -62,18 +62,25 @@ def add(self, blocks: list[Block] | Block, fail_on_duplicate_key: bool = True):
 
     def _find_duplicate_keys(self, blocks: list[Block]) -> list[str]:
         """Keys of blocks that would become duplicates when added to the library."""
+        # Look keys up in the by-key dicts directly (rather than copying them into
+        # sets), so that the cost depends only on the number of blocks being added,
+        # not on the size of the library.
         duplicate_keys = []
-        seen_entry_keys = set(self._entries_by_key)
-        seen_string_keys = set(self._strings_by_key)
+        new_entry_keys = set()
+        new_string_keys = set()
         for block in blocks:
             if isinstance(block, Entry):
-                if block.key in seen_entry_keys:
-                    duplicate_keys.append(block.key)
-                seen_entry_keys.add(block.key)
+                key = block.key
+                if key in self._entries_by_key or key in new_entry_keys:
+                    duplicate_keys.append(key)
+                else:
+                    new_entry_keys.add(key)
             elif isinstance(block, String):
-                if block.key in seen_string_keys:
-                    duplicate_keys.append(block.key)
-                seen_string_keys.add(block.key)
+                key = block.key
+                if key in self._strings_by_key or key in new_string_keys:
+                    duplicate_keys.append(key)
+                else:
+                    new_string_keys.add(key)
         return duplicate_keys
 
     def _block_index(self, block: Block, skip: set[int] | None = None) -> int:

bibtexparser/middlewares/enclosing.py

@@ -121,8 +121,11 @@ def _is_enclosed_in_quotes(value: str) -> bool:
         concatenation expression `"intro" # "outro"`. Escaped quotes and quotes
         inside braces are not counted.
         """
-        depth = 0
         last_index = len(value) - 1
+        if value.find('"', 1, last_index) < 0:
+            # Fast path for the common case of a value without inner quotes.
+            return True
+        depth = 0
         for match in _BRACES_AND_QUOTE.finditer(value):
             index = match.start()
             if index == 0 or index == last_index:

bibtexparser/splitter.py

@@ -308,16 +308,16 @@ def split(self) -> Library:
                 # Clean up previous block implicit_comment
                 implicit_comment = self._end_implicit_comment(m.start())
                 if implicit_comment is not None:
-                    library.add(implicit_comment)
+                    library.add(implicit_comment, fail_on_duplicate_key=False)
                 self._implicit_comment_start = None
 
                 start_line = self._current_line
                 try:
                     # Start new block parsing
                     if m_val.startswith("@comment"):
-                        library.add(self._handle_explicit_comment())
+                        library.add(self._handle_explicit_comment(), fail_on_duplicate_key=False)
                     elif m_val.startswith("@preamble"):
-                        library.add(self._handle_preamble())
+                        library.add(self._handle_preamble(), fail_on_duplicate_key=False)
                     elif m_val.startswith("@string"):
                         library.add(self._handle_string(m), fail_on_duplicate_key=False)
                     else:
@@ -338,7 +338,8 @@ def split(self) -> Library:
                             start_line=start_line,
                             raw=self.bibstr[m.start() : e.end_index],
                             error=e,
-                        )
+                        ),
+                        fail_on_duplicate_key=False,
                     )
 
                 except ParserStateException as e:
@@ -365,7 +366,7 @@ def split(self) -> Library:
         if self._implicit_comment_start is not None:
             comment = self._end_implicit_comment(len(self.bibstr))
             if comment is not None:
-                library.add(comment)
+                library.add(comment, fail_on_duplicate_key=False)
 
         return library
 

tests/test_library.py

@@ -218,3 +218,37 @@ def test_replace_entry_with_changed_key():
     assert library.blocks == [replacement]
     assert library.entries == [replacement]
     assert library.entries_dict == {"replacementKey": replacement}
+
+
+class _NonIterableDict(dict):
+    """A dict that supports lookups but refuses to be iterated or copied.
+
+    Used to make sure that `Library.add` does not walk the whole by-key dicts
+    (i.e., does not scale with the library size) when checking for duplicates."""
+
+    def __iter__(self):
+        raise AssertionError("by-key dict must not be iterated on add")
+
+    def keys(self):
+        raise AssertionError("by-key dict must not be iterated on add")
+
+
+def test_add_duplicate_check_does_not_iterate_library():
+    """Checking for duplicates must not iterate over all existing keys (was O(library size))."""
+    library = Library()
+    library.add([get_dummy_entry_with_key("existing"), String(key="s", value="{v}")])
+    library._entries_by_key = _NonIterableDict(library._entries_by_key)
+    library._strings_by_key = _NonIterableDict(library._strings_by_key)
+
+    library.add(ImplicitComment(comment="a comment"))
+    library.add(get_dummy_entry_with_key("new"))
+    library.add(String(key="t", value="{w}"))
+    assert len(library.blocks) == 5
+
+    with pytest.raises(ValueError):
+        library.add(get_dummy_entry_with_key("existing"))
+    with pytest.raises(ValueError):
+        library.add(String(key="s", value="{v}"))
+    with pytest.raises(ValueError):
+        library.add([get_dummy_entry_with_key("twice"), get_dummy_entry_with_key("twice")])
+    assert len(library.blocks) == 5