Filter by property equal not working as expected on string

Support
1

Description

Filtering by equal on a text property seems to result in a “contains” filter instead.

Example:

from weaviate.classes.query import Filter
from weaviate.connect import ConnectionParams
from weaviate import WeaviateClient

weaviate_client = WeaviateClient(
    connection_params=ConnectionParams.from_url(      
        os.getenv("WEAVIATE_SERVER_URL","http://localhost:8080"),  
        int(os.getenv("WEAVIATE_SERVER_GRPC_PORT", 50051))    
))

col = weaviate_client.collections.use(index)
response = col.query.fetch_objects(filters=Filter.by_property(“drillDiameter”).equal(“2 mm”),limit=3)

for o in response.objects:
    print(o.properties[‘drillDiameter’])

outputs:

2,4 mm
2,1 mm
5,2 mm

Server Setup Information

  • Weaviate Server Version: 1.30.2
  • Deployment Method: local Docker run semitechnologies/weaviate:1.30.2
  • Multi Node? Number of Running Nodes: default settings
  • Client Language and Version: weaviate-client 4.16.10 (python, poetry-installed)
  • Multitenancy?: default settings

Is this a known bug or may it be related to my specific test case or setup? I couldn’t find anything about this yet, any help would be much appreciated.

2

Hi!
The tokenization property has an impact on the filtering process. What tokenization are you using?
You can read more about it here - Tokenization and filters | Weaviate Documentation

2 Likes
3

That’s right! Thanks @Bigdwarf43 !

So when you index 2 mm using word tokenization (default) you end up with two tokens: 2 and mm.

If you set the tokenization to field, you will then get a token with 2 mm

Happy coding!

1 Like