Issues facing while performing CRUD operations

Support
,
1

I am facing an issue with deleting and retrieving data when I have applied where filter.

I want to delete only the name test, but it is deleting the test and test limi properties/data objects I have in my weaviate. I don’t want this to happen.
Here, if you observe that the display or get function deviates from retrieving results based on semantic search, but I need to extract search results. For example, for a test, I need only matches of the test, not the test limi, etc.

My query::
check_query = f"“”
{{
Get {{
{class_name}(
where: {{
path: [“name”],
operator: Equal,
valueString: “{name}”
}}
) {{
name
_additional {{
id
}}
}}
}}
}}
“”"
response = client.query.raw(check_query)

print(“Response from Weaviate:”, response)
delete_query = {
“operator”: “Equal”,
“path”: [“name”],
“valueString”: name,
}

x= client.batch.delete_objects(
class_name=class_name,
where=delete_query
)
print(x)
Response from Weaviate: {‘data’: {‘Get’: {‘Issue’: [{‘_additional’: {‘id’: ‘a5fff400-df0f-4320-8134-036ae28c4891’}, ‘name’: ‘test limi’}, {‘_additional’: {‘id’: ‘9cb804bc-8375-4dfe-bf4d-f30402780386’}, ‘name’: ‘test’}]}}}

Delete query response :

{‘dryRun’: False, ‘match’: {‘class’: ‘Issue’, ‘where’: {‘operands’: None, ‘operator’: ‘Equal’, ‘path’: [‘name’], ‘valueString’: ‘test’}}, ‘output’: ‘minimal’, ‘results’: {‘failed’: 0, ‘limit’: 10000, ‘matches’: 2, ‘objects’: None, ‘successful’: 2}}

Successfully deleted 4 exact matches for client name ‘test’

Support

Resolve this issue if anyone aware of
I am using docker version of weavaite and my tokenizer for this property is whitespace

2

Hi!

What is the server version?

You property probably has the tokenization of the property name set to field.

This means that My Cool Test will become three tokens: my cool test, and now if you search for some objects that has cool it will find all objects that has this token.

Check here for a comprehensive doc on Tokenization in Weaviate:

You may want to set the tokenization to field to avoid this scenario.

Let me know if that helps!

Thanks!

3

No, I have used “white space” tokenization in my schema while creating a class.

And I am using python client v3

4

That best thing here is to create some python code, if possible in pyv4, so we can try to reproduce the issue you are facing.

Can you share a code for this?

Thanks!