jbellis commented on code in PR #12373:
URL: https://github.com/apache/lucene/pull/12373#discussion_r1237022527
##########
lucene/core/src/java/org/apache/lucene/index/FloatVectorValues.java:
##########
@@ -31,6 +31,15 @@ public abstract class FloatVectorValues extends
DocIdSetIterator {
/** The maximum length of a vector */
public static final int MAX_DIMENSIONS = 1024;
+ /**
+ * This is the largest float vector value that we allow.
+ *
+ * <p>The largest float32 that you can square without overflowing is about
1.8E19. We reduce that
+ * further to accommodate the addition of multiple such components in
similarity computations in
+ * vectors up to MAX_DIMENSIONS in length.
+ */
+ public static final float MAX_FLOAT32_COMPONENT = 1E17f;
Review Comment:
that's true, but there's also something to be said for "here's a simple rule
to follow" vs "look up the docs on every distance function to see what the
limit is"
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
