Uploaded image for project: 'MariaDB Server'
  1. MariaDB Server
  2. MDEV-32885

VEC_DISTANCE() function

Details

    Description

      VEC_DISTANCE() function that takes two binary strings (which are vectors of floats of the same length) and computes Euclidean distance between two multi-dimensional points, much like ST_DISTANCE() does.

      it'll also work on the VECTOR(N) data type, when we'll have one

      Attachments

        Issue Links

          is part of

          New Feature - A new feature of the product, which has yet to be developed. MDEV-34939 vector search in 11.7

          • Critical - Crashes, loss of data, severe memory leak.
          • Closed
          relates to

          New Feature - A new feature of the product, which has yet to be developed. MDEV-32886 VEC_FromText() and VEC_ToText() functions

          • Major - Major loss of function.
          • Closed

          Epic - Created by Jira Software - do not edit or delete. Issue type for a big user story that needs to be broken down. MDEV-32887 vector search

          • Critical - Crashes, loss of data, severe memory leak.
          • Stalled

          New Feature - A new feature of the product, which has yet to be developed. MDEV-33417 VEC_DISTANCE_COSINE() function

          • Major - Major loss of function.
          • Closed

          New Feature - A new feature of the product, which has yet to be developed. MDEV-33410 VECTOR data type

          • Critical - Crashes, loss of data, severe memory leak.
          • Closed

          New Feature - A new feature of the product, which has yet to be developed. MDEV-35327 Add VEC_DISTANCE_MANHATTAN (L1) distance function on Vector

          • Minor - Minor loss of function, or other problem where easy workaround is present.
          • Open

          Activity

            Ascending order - Click to sort in descending order
            piki Patrick Reynolds added a comment -

            PlanetScale intends to support three or more different distance functions, so we need a syntax for that. Our proposal is VEC_L2Distance, VEC_IPDistance, and VEC_CosineDistance. This is similar to spatial types, which have separate functions for simple, Hausdorff, and Frechet distance.

            We may in the future also want VEC_HammingDistance for one-bit vectors.

            piki Patrick Reynolds added a comment - PlanetScale intends to support three or more different distance functions, so we need a syntax for that. Our proposal is VEC_L2Distance, VEC_IPDistance, and VEC_CosineDistance. This is similar to spatial types, which have separate functions for simple, Hausdorff, and Frechet distance. We may in the future also want VEC_HammingDistance for one-bit vectors.
            rdyas Robert Dyas added a comment -

            Presumably all vector distance functions return a single float between 0 and 1 with 1.0 being a perfect match.

            rdyas Robert Dyas added a comment - Presumably all vector distance functions return a single float between 0 and 1 with 1.0 being a perfect match.
            serg Sergei Golubchik added a comment - - edited

            No, not at all. If the distance is L2 (euclidean distance, |x - y|, then the result is a non-negative number, where 0 means a perfect match. That's how the euclidean distance is defined.

            If the distance is cosine (1-cos(∠xy)), then it's 0 being a perfect match, 2 being the farthest possible. Again, that's just what a cosine distance is.

            serg Sergei Golubchik added a comment - - edited No, not at all. If the distance is L2 (euclidean distance, | x - y | , then the result is a non-negative number, where 0 means a perfect match. That's how the euclidean distance is defined. If the distance is cosine ( 1-cos(∠ xy ) ), then it's 0 being a perfect match, 2 being the farthest possible. Again, that's just what a cosine distance is.
            rdyas Robert Dyas added a comment -

            Thank you for the clarification.

            rdyas Robert Dyas added a comment - Thank you for the clarification.
            serg Sergei Golubchik added a comment -

            as we're modelling function names after GIS, we'll use ST_DISTANCE() and ST_DISTANCE_SPHERE() as role models. Meaning we'll have VEC_DISTANCE_EUCLIDEAN(), VEC_DISTANCE_COSINE(), etc.

            May be VEC_DISTANCE() should mean Euclidean, but it'll likely be too confusing.

            serg Sergei Golubchik added a comment - as we're modelling function names after GIS, we'll use ST_DISTANCE() and ST_DISTANCE_SPHERE() as role models. Meaning we'll have VEC_DISTANCE_EUCLIDEAN() , VEC_DISTANCE_COSINE() , etc. May be VEC_DISTANCE() should mean Euclidean, but it'll likely be too confusing.

            People

              serg Sergei Golubchik
              serg Sergei Golubchik
              Votes:
              2 Vote for this issue
              Watchers:
              19 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved:

                Git Integration

                  Error rendering 'com.xiplink.jira.git.jira_git_plugin:git-issue-webpanel'. Please contact your Jira administrators.