René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"HnswDocumentIndex treats document IDs as string, they can be str, int, ID","articleBody":"### Initial Checks\n\n- [X] I have read and followed [the docs](https://docs.docarray.org/) and still think this is a bug\n\n### Description\n\nI noticed this behavior when I wanted to access multiple documents in the index:\r\n\r\n```python\r\n@requests(on='/find')\r\n    def find(self, docs: DocList[QuoteFile], **_) -\u003e DocList[QuoteFile]:\r\n        return self._cache_di[docs.id]\r\n```\r\n\r\nAnd when I issue POST /find with body `{\"data\":[{\"id\":\"300055\"}]}`, this code yields:\r\n```bash\r\n       \"/Users/oytuntez/motaword/jina-documents/venv/lib/py…                    \r\n       line 544, in _get_docs_sqlite_doc_id                                     \r\n           hashed_ids = tuple(self._to_hashed_id(id_) for                       \r\n       id_ in doc_ids)                                                          \r\n         File                                                                   \r\n       \"/Users/oytuntez/motaword/jina-documents/venv/lib/py…                    \r\n       line 544, in \u003cgenexpr\u003e                                                   \r\n           hashed_ids = tuple(self._to_hashed_id(id_) for                       \r\n       id_ in doc_ids)                                                          \r\n         File                                                                   \r\n       \"/Users/oytuntez/motaword/jina-documents/venv/lib/py…                    \r\n       line 445, in _to_hashed_id                                               \r\n           return                                                               \r\n       int(hashlib.sha256(doc_id.encode('utf-8')).hexdigest…                    \r\n       16) % 10**18                                                             \r\n       AttributeError: 'int' object has no attribute                            \r\n       'encode'                        \r\n```\r\n\r\nUpon investigation, I saw that most of `HnswDocumentIndex` treats IDs as `str`. However, it is my understanding that IDs can be `int`, see this type definition:\r\n\r\n```python\r\nclass ID(str, AbstractType):\r\n    \"\"\"\r\n    Represent an unique ID\r\n    \"\"\"\r\n\r\n    @classmethod\r\n    def _docarray_validate(\r\n        cls: Type[T],\r\n        value: Union[str, int, UUID],\r\n...\r\n```\r\n\r\nI think ID values should be cast to `str` if necessary (it would be in `_to_hashed_id` case).\n\n### Example Code\n\n_No response_\n\n### Python, DocArray \u0026 OS Version\n\n```Text\nPython 3.8.12\r\ndocarray==0.40.0\n```\n\n\n### Affected Components\n\n- [X] [Vector Database / Index](https://docs.docarray.org/user_guide/storing/docindex/)\n- [ ] [Representing](https://docs.docarray.org/user_guide/representing/first_step)\n- [ ] [Sending](https://docs.docarray.org/user_guide/sending/first_step/)\n- [X] [storing](https://docs.docarray.org/user_guide/storing/first_step/)\n- [X] [multi modal data type](https://docs.docarray.org/data_types/first_steps/)","author":{"url":"https://github.com/oytuntez","@type":"Person","name":"oytuntez"},"datePublished":"2024-02-07T23:23:17.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":6},"url":"https://github.com/1850/docarray/issues/1850"}