You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Andres de la Peña (Jira)" <ji...@apache.org> on 2023/06/20 18:06:00 UTC
[jira] [Comment Edited] (CASSANDRA-18613) Add support for vectors on UDFs
[ https://issues.apache.org/jira/browse/CASSANDRA-18613?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17735403#comment-17735403 ]
Andres de la Peña edited comment on CASSANDRA-18613 at 6/20/23 6:05 PM:
------------------------------------------------------------------------
Just updated the patch for supporting vectors of variable-length elements, such as {{{}varint{}}}, {{{}text{}}}, etc. The patch is [here|https://github.com/adelapena/cassandra/commit/07e11b091fa3df03a10913e8a101c76df3f7796f].
By the way, I think that vectors of variable-length elements are [not currently supported by the Java driver|https://github.com/datastax/java-driver/blob/cfeb55f8ba9c35bd3b19c9442b429a564b565440/core/src/main/java/com/datastax/oss/driver/internal/core/type/codec/CqlVectorCodec.java#L91-L95].
The driver uses a dedicated [{{CqlVector}}|https://github.com/datastax/java-driver/blob/cfeb55f8ba9c35bd3b19c9442b429a564b565440/core/src/main/java/com/datastax/oss/driver/api/core/data/CqlVector.java] class to represent vector values. However, for the UDF codec I have preferred to just represent vectors as Java lists. The reason is that I think that it's easier for users writing a UDF to simply generate new vector values as instances of a standard Java class. Using a non-standard class can be a bit cumbersome when trying to fit code in the restricted environment of a UDF, and it doesn't seem to provide any additional value.
was (Author: adelapena):
Just updated the patch for supporting vectors of variable-length elements, such as {{{}varint{}}}, {{{}text{}}}, etc. The patch is [here|https://github.com/adelapena/cassandra/commit/07e11b091fa3df03a10913e8a101c76df3f7796f].
By the way, I think that vectors of variable-length elements are [not currently supported by the Java driver|https://github.com/datastax/java-driver/blob/cfeb55f8ba9c35bd3b19c9442b429a564b565440/core/src/main/java/com/datastax/oss/driver/internal/core/type/codec/CqlVectorCodec.java#L91-L95].
The driver uses a dedicated [{{CqlVector}}|https://github.com/datastax/java-driver/blob/cfeb55f8ba9c35bd3b19c9442b429a564b565440/core/src/main/java/com/datastax/oss/driver/api/core/data/CqlVector.java] class to represent vector values. However, for the UDF codec I have preferred to just represent vectors as a list. The reason is that I think that it's easier for users writing a UDF to simply generate new vector values as standard Java lists. Using a non-standard class can be a bit cumbersome when trying to fit code in the restricted environment of a UDF, and it doesn't seem to provide any additional value.
> Add support for vectors on UDFs
> -------------------------------
>
> Key: CASSANDRA-18613
> URL: https://issues.apache.org/jira/browse/CASSANDRA-18613
> Project: Cassandra
> Issue Type: New Feature
> Components: Cluster/Schema
> Reporter: Andres de la Peña
> Assignee: Andres de la Peña
> Priority: Normal
> Fix For: 5.x
>
>
> CASSANDRA-18504 will add a new vector type, but [it won't be supported on UDFs|https://github.com/apache/cassandra/blob/5027e688da006e5d5bf9bfdf4719caddbf85986a/test/unit/org/apache/cassandra/cql3/validation/operations/CQLVectorTest.java#L248-L271]. The goal of this ticket is to add that support.
> This will require adding a new {{o.a.c.cql3.functions.types.TypeCodec}} for vectors. Those codecs are mostly duplicates of the codecs on the Java driver. They are used for UDFs instead of the regular {{AbstractType}} to prevent pulling too many internal dependencies. The driver's vector codec has recently been added by [JAVA-3060|https://datastax-oss.atlassian.net/browse/JAVA-3060].
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@cassandra.apache.org
For additional commands, e-mail: commits-help@cassandra.apache.org